Search results for: machine intelligence

Authors: Payam Mousavi, Andrew L. Stewart, Huiwen You, Aryeh F. G. Fayerman

Abstract:

We present a decision-support tool to assist an operator in the detection and tracking of a suspect vehicle traveling to an unknown target destination. Multiple data sources, such as traffic cameras, traffic information, weather, etc., are integrated and processed in real-time to infer a suspect’s intended destination chosen from a list of pre-determined high-value targets. Previously, we presented our work in the detection and tracking of vehicles using traffic and airborne cameras. Here, we focus on the fusion and processing of that information to predict a suspect’s behavior. The network of cameras is represented by a directional graph, where the edges correspond to direct road connections between the nodes and the edge weights are proportional to the average time it takes to travel from one node to another. For our experiments, we construct our graph based on the greater Los Angeles subset of the Caltrans’s “Performance Measurement System” (PeMS) dataset. We propose a Bayesian approach where a posterior probability for each target is continuously updated based on detections of the suspect in the live video feeds. Additionally, we introduce the concept of ‘soft interventions’, inspired by the field of Causal Inference. Soft interventions are herein defined as interventions that do not immediately interfere with the suspect’s movements; rather, a soft intervention may induce the suspect into making a new decision, ultimately making their intent more transparent. For example, a soft intervention could be temporarily closing a road a few blocks from the suspect’s current location, which may require the suspect to change their current course. The objective of these interventions is to gain the maximum amount of information about the suspect’s intent in the shortest possible time. Our system currently operates in a human-on-the-loop mode where at each step, a set of recommendations are presented to the operator to aid in decision-making. In principle, the system could operate autonomously, only prompting the operator for critical decisions, allowing the system to significantly scale up to larger areas and multiple suspects. Once the intended target is identified with sufficient confidence, the vehicle is reported to the authorities to take further action. Other recommendations include a selection of road closures, i.e., soft interventions, or to continue monitoring. We evaluate the performance of the proposed system using simulated scenarios where the suspect, starting at random locations, takes a noisy shortest path to their intended target. In all scenarios, the suspect’s intended target is unknown to our system. The decision thresholds are selected to maximize the chances of determining the suspect’s intended target in the minimum amount of time and with the smallest number of interventions. We conclude by discussing the limitations of our current approach to motivate a machine learning approach, based on reinforcement learning in order to relax some of the current limiting assumptions.

Keywords: autonomous surveillance, Bayesian reasoning, decision support, interventions, patterns of life, predictive analytics, predictive insights

Procedia PDF Downloads 95

Authors: Shayan Mohajer Hamidi

Abstract:

Federated Learning (FL) has emerged as a promising paradigm for training machine learning models on distributed data without the need for centralized data aggregation. In the realm of wireless communications, FL has the potential to leverage the vast amounts of data generated by wireless devices to improve model performance and enable intelligent applications. However, the fairness aspect of FL in wireless communications remains largely unexplored. This abstract presents an idea for fair federated learning in wireless communications, addressing the challenges of imbalanced data distribution, privacy preservation, and resource allocation. Firstly, the proposed approach aims to tackle the issue of imbalanced data distribution in wireless networks. In typical FL scenarios, the distribution of data across wireless devices can be highly skewed, resulting in unfair model updates. To address this, we propose a weighted aggregation strategy that assigns higher importance to devices with fewer samples during the aggregation process. By incorporating fairness-aware weighting mechanisms, the proposed approach ensures that each participating device's contribution is proportional to its data distribution, thereby mitigating the impact of data imbalance on model performance. Secondly, privacy preservation is a critical concern in federated learning, especially in wireless communications where sensitive user data is involved. The proposed approach incorporates privacy-enhancing techniques, such as differential privacy, to protect user privacy during the model training process. By adding carefully calibrated noise to the gradient updates, the proposed approach ensures that the privacy of individual devices is preserved without compromising the overall model accuracy. Moreover, the approach considers the heterogeneity of devices in terms of computational capabilities and energy constraints, allowing devices to adaptively adjust the level of privacy preservation to strike a balance between privacy and utility. Thirdly, efficient resource allocation is crucial for federated learning in wireless communications, as devices operate under limited bandwidth, energy, and computational resources. The proposed approach leverages optimization techniques to allocate resources effectively among the participating devices, considering factors such as data quality, network conditions, and device capabilities. By intelligently distributing the computational load, communication bandwidth, and energy consumption, the proposed approach minimizes resource wastage and ensures a fair and efficient FL process in wireless networks. To evaluate the performance of the proposed fair federated learning approach, extensive simulations and experiments will be conducted. The experiments will involve a diverse set of wireless devices, ranging from smartphones to Internet of Things (IoT) devices, operating in various scenarios with different data distributions and network conditions. The evaluation metrics will include model accuracy, fairness measures, privacy preservation, and resource utilization. The expected outcomes of this research include improved model performance, fair allocation of resources, enhanced privacy preservation, and a better understanding of the challenges and solutions for fair federated learning in wireless communications. The proposed approach has the potential to revolutionize wireless communication systems by enabling intelligent applications while addressing fairness concerns and preserving user privacy.

Keywords: federated learning, wireless communications, fairness, imbalanced data, privacy preservation, resource allocation, differential privacy, optimization

Procedia PDF Downloads 51

Authors: Shanna Swart, Caryn Fenner, Athanasios Kotsiopoulos, Susan Harrison

Abstract:

Linear alkanes are produced as by-products from the increasing use of gas-to-liquid fuel technologies for synthetic fuel production and offer great potential for value addition. Their current use as low-value fuels and solvents do not maximize this potential. Therefore, attention has been drawn towards direct activation of these aliphatic alkanes to more useful products such as alcohols, aldehydes, carboxylic acids and derivatives. Cytochrome P450 monooxygenases (P450s) can be used for activation of these aliphatic alkanes using whole-cells or cell-free systems. Some limitations of whole-cell systems include reduced mass transfer, stability and possible side reactions. Since the P450 systems are little studied as cell-free systems, they form the focus of this study. Challenges of a cell-free system include co-factor regeneration, substrate availability and enzyme stability. Enzyme immobilization offers a positive outlook on this dilemma, as it may enhance stability of the enzyme. In the present study, 2 different P450s (CYP153A6 and CYP102A1) as well as the relevant accessory enzymes required for electron transfer (ferredoxin and ferredoxin reductase) and co-factor regeneration (glucose dehydrogenase) have been expressed in E. coli and purified by metal affinity chromatography. Glucose dehydrogenase (GDH), was used as a model enzyme to assess the potential of various enzyme immobilization strategies including; surface attachment on MagReSyn® microspheres with various functionalities and on electrospun nanofibers, using self-assembly based methods forming Cross Linked Enzymes (CLE), Cross Linked Enzyme Aggregates (CLEAs) and spherezymes as well as in a sol gel. The nanofibers were synthesized by electrospinning, which required the building of an electrospinning machine. The nanofiber morphology has been analyzed by SEM and binding will be further verified by FT-IR. Covalent attachment based methods showed limitations where only ferredoxin reductase and GDH retained activity after immobilization which were largely attributed to insufficient electron transfer and inactivation caused by the crosslinkers (60% and 90% relative activity loss for the free enzyme when using 0.5% glutaraldehyde and glutaraldehyde/ethylenediamine (1:1 v/v), respectively). So far, initial experiments with GDH have shown the most potential when immobilized via their His-tag onto the surface of MagReSyn® microspheres functionalized with Ni-NTA. It was found that Crude GDH could be simultaneously purified and immobilized with sufficient activity retention. Immobilized pure and crude GDH could be recycled 9 and 10 times, respectively, with approximately 10% activity remaining. The immobilized GDH was also more stable than the free enzyme after storage for 14 days at 4˚C. This immobilization strategy will also be applied to the P450s and optimized with regards to enzyme loading and immobilization time, as well as characterized and compared with the free enzymes. It is anticipated that the proposed immobilization set-up will offer enhanced enzyme stability (as well as reusability and easy recovery), minimal mass transfer limitation, with continuous co-factor regeneration and minimal enzyme leaching. All of which provide a positive outlook on this robust multi-enzyme system for efficient activation of linear alkanes as well as the potential for immobilization of various multiple enzymes, including multimeric enzymes for different bio-catalytic applications beyond alkane activation.

Keywords: alkane activation, cytochrome P450 monooxygenase, enzyme catalysis, enzyme immobilization

Procedia PDF Downloads 204

Authors: Ho-Won Jung, Donghun Lee, Hyung-Jin Kim

Abstract:

Predictive analytics (data analysis) uses a subset of measurements (the features, predictor, or independent variable) to predict another measurement (the outcome, target, or dependent variable) on a single person or unit. It applies empirical methods in statistics, operations research, and machine learning to predict the future, or otherwise unknown events or outcome on a single or person or unit, based on patterns in data. Most analyses of metabolic syndrome are not predictive analytics but statistical explanatory studies that build a proposed model (theory building) and then validate metabolic syndrome predictors hypothesized (theory testing). A proposed theoretical model forms with causal hypotheses that specify how and why certain empirical phenomena occur. Predictive analytics and explanatory modeling have their own territories in analysis. However, predictive analytics can perform vital roles in explanatory studies, i.e., scientific activities such as theory building, theory testing, and relevance assessment. In the context, this study is to demonstrate how to use our predictive analytics to support theory building (i.e., hypothesis generation). For the purpose, this study utilized a big data predictive analytics platform TM based on a co-occurrence graph. The co-occurrence graph is depicted with nodes (e.g., items in a basket) and arcs (direct connections between two nodes), where items in a basket are fully connected. A cluster is a collection of fully connected items, where the specific group of items has co-occurred in several rows in a data set. Clusters can be ranked using importance metrics, such as node size (number of items), frequency, surprise (observed frequency vs. expected), among others. The size of a graph can be represented by the numbers of nodes and arcs. Since the size of a co-occurrence graph does not depend directly on the number of observations (transactions), huge amounts of transactions can be represented and processed efficiently. For a demonstration, a total of 13,254 metabolic syndrome training data is plugged into the analytics platform to generate rules (potential hypotheses). Each observation includes 31 predictors, for example, associated with sociodemographic, habits, and activities. Some are intentionally included to get predictive analytics insights on variable selection such as cancer examination, house type, and vaccination. The platform automatically generates plausible hypotheses (rules) without statistical modeling. Then the rules are validated with an external testing dataset including 4,090 observations. Results as a kind of inductive reasoning show potential hypotheses extracted as a set of association rules. Most statistical models generate just one estimated equation. On the other hand, a set of rules (many estimated equations from a statistical perspective) in this study may imply heterogeneity in a population (i.e., different subpopulations with unique features are aggregated). Next step of theory development, i.e., theory testing, statistically tests whether a proposed theoretical model is a plausible explanation of a phenomenon interested in. If hypotheses generated are tested statistically with several thousand observations, most of the variables will become significant as the p-values approach zero. Thus, theory validation needs statistical methods utilizing a part of observations such as bootstrap resampling with an appropriate sample size.

Keywords: explanatory modeling, metabolic syndrome, predictive analytics, theory building

Procedia PDF Downloads 248

Authors: D. Madhushanka, Y. Liu, H. C. Fernando

Abstract:

Fires are one of the foremost factors of land surface disturbance in diverse ecosystems, causing soil erosion and land-cover changes and atmospheric effects affecting people's lives and properties. Generally, the severity of the fire is calculated as the Normalized Burn Ratio (NBR) index. This is performed manually by comparing two images obtained afterward. Then by using the bitemporal difference of the preprocessed satellite images, the dNBR is calculated. The burnt area is then classified as either unburnt (dNBR<0.1) or burnt (dNBR>= 0.1). Furthermore, Wildfire Severity Assessment (WSA) classifies burnt areas and unburnt areas using classification levels proposed by USGS and comprises seven classes. This procedure generates a burn severity report for the area chosen by the user manually. This study is carried out with the objective of producing an automated tool for the above-mentioned process, namely the World Wildfire Severity Assessment Tool (WWSAT). It is implemented in Google Earth Engine (GEE), which is a free cloud-computing platform for satellite data processing, with several data catalogs at different resolutions (notably Landsat, Sentinel-2, and MODIS) and planetary-scale analysis capabilities. Sentinel-2 MSI is chosen to obtain regular processes related to burnt area severity mapping using a medium spatial resolution sensor (15m). This tool uses machine learning classification techniques to identify burnt areas using NBR and to classify their severity over the user-selected extent and period automatically. Cloud coverage is one of the biggest concerns when fire severity mapping is performed. In WWSAT based on GEE, we present a fully automatic workflow to aggregate cloud-free Sentinel-2 images for both pre-fire and post-fire image compositing. The parallel processing capabilities and preloaded geospatial datasets of GEE facilitated the production of this tool. This tool consists of a Graphical User Interface (GUI) to make it user-friendly. The advantage of this tool is the ability to obtain burn area severity over a large extent and more extended temporal periods. Two case studies were carried out to demonstrate the performance of this tool. The Blue Mountain national park forest affected by the Australian fire season between 2019 and 2020 is used to describe the workflow of the WWSAT. This site detected more than 7809 km2, using Sentinel-2 data, giving an error below 6.5% when compared with the area detected on the field. Furthermore, 86.77% of the detected area was recognized as fully burnt out, of which high severity (17.29%), moderate-high severity (19.63%), moderate-low severity (22.35%), and low severity (27.51%). The Arapaho and Roosevelt National Forest Park, California, the USA, which is affected by the Cameron peak fire in 2020, is chosen for the second case study. It was found that around 983 km2 had burned out, of which high severity (2.73%), moderate-high severity (1.57%), moderate-low severity (1.18%), and low severity (5.45%). These spots also can be detected through the visual inspection made possible by cloud-free images generated by WWSAT. This tool is cost-effective in calculating the burnt area since satellite images are free and the cost of field surveys is avoided.

Keywords: burnt area, burnt severity, fires, google earth engine (GEE), sentinel-2

Procedia PDF Downloads 206

Authors: David Bakker, Nikki Rickard

Abstract:

Smartphone apps hold immense potential as mental health and wellbeing tools. Support can be made easily accessible and can be used in real-time while users are experiencing distress. Furthermore, data can be collected to enable machine learning and automated tailoring of support to users. While many apps have been developed for mental health purposes, few have adhered to evidence-based recommendations and even fewer have pursued experimental validation. This paper details the development and experimental evaluation of an app, MoodMission, that aims to provide support for low moods and anxiety, help prevent clinical depression and anxiety disorders, and serve as an adjunct to professional clinical supports. MoodMission was designed to deliver cognitive behavioural therapy for specifically reported problems in real-time, momentary interactions. Users report their low moods or anxious feelings to the app along with a subjective units of distress scale (SUDS) rating. MoodMission then provides a choice of 5-10 short, evidence-based mental health strategies called Missions. Users choose a Mission, complete it, and report their distress again. Automated tailoring, gamification, and in-built data collection for analysis of effectiveness was also included in the app’s design. The development process involved construction of an evidence-based behavioural plan, designing of the app, building and testing procedures, feedback-informed changes, and a public launch. A randomized controlled trial (RCT) was conducted comparing MoodMission to two other apps and a waitlist control condition. Participants completed measures of anxiety, depression, well-being, emotional self-awareness, coping self-efficacy and mental health literacy at the start of their app use and 30 days later. At the time of submission (November 2016) over 300 participants have participated in the RCT. Data analysis will begin in January 2017. At the time of this submission, MoodMission has over 4000 users. A repeated-measures ANOVA of 1390 completed Missions reveals that SUDS (0-10) ratings were significantly reduced between pre-Mission ratings (M=6.20, SD=2.39) and post-Mission ratings (M=4.93, SD=2.25), F(1,1389)=585.86, p < .001, np2=.30. This effect was consistent across both low moods and anxiety. Preliminary analyses of the data from the outcome measures surveys reveal improvements across mental health and wellbeing measures as a result of using the app over 30 days. This includes a significant increase in coping self-efficacy, F(1,22)=5.91, p=.024, np2=.21. Complete results from the RCT in which MoodMission was evaluated will be presented. Results will also be presented from the continuous outcome data being recorded by MoodMission. MoodMission was successfully developed and launched, and preliminary analysis suggest that it is an effective mental health and wellbeing tool. In addition to the clinical applications of MoodMission, the app holds promise as a research tool to conduct component analysis of psychological therapies and overcome restraints of laboratory based studies. The support provided by the app is discrete, tailored, evidence-based, and transcends barriers of stigma, geographic isolation, financial limitations, and low health literacy.

Keywords: anxiety, app, CBT, cognitive behavioural therapy, depression, eHealth, mission, mobile, mood, MoodMission

Procedia PDF Downloads 251

Authors: Emre Kara, Ali Kurşun, Halil Aykul

Abstract:

The lightweight sandwiches obtained with the use of various core materials such as foams, honeycomb, lattice structures etc., which have high energy absorbing capacity and high strength to weight ratio, are suitable for several applications in transport industry (automotive, aerospace, shipbuilding industry) where saving of fuel consumption, load carrying capacity increase, safety of vehicles and decrease of emission of harmful gases are very important aspects. While the sandwich structures with foams and honeycombs have been applied for many years, there is a growing interest on a new generation sandwiches with micro lattice cores. In order to produce these core structures, various production methods were created with the development of the technology. One of these production technologies is an additive manufacturing technique called selective laser sintering/melting (SLS/SLM) which is very popular nowadays because of saving of production time and achieving the production of complex topologies. The static bending and the dynamic low velocity impact tests of the sandwiches with carbon fiber/epoxy skins and the micro lattice cores produced via SLS/SLM were already reported in just a few studies. The goal of this investigation was the analysis of the flexural response of the sandwiches consisting of glass fiber reinforced plastic (GFRP) skins and the micro lattice cores manufactured via SLS under thermo-mechanical loads in order to compare the results in terms of peak load and absorbed energy values respect to the effect of core cell size, temperature and support span length. The micro lattice cores were manufactured using SLS technology that creates the product drawn by a 3D computer aided design (CAD) software. The lattice cores which were designed as body centered cubic (BCC) model having two different cell sizes (d= 2 and 2.5 mm) with the strut diameter of 0.3 mm were produced using titanium alloy (Ti6Al4V) powder. During the production of all the core materials, the same production parameters such as laser power, laser beam diameter, building direction etc. were kept constant. Vacuum Infusion (VI) method was used to produce skin materials, made of [0°/90°] woven S-Glass prepreg laminates. The combination of the core and skins were implemented under VI. Three point bending tests were carried out by a servo-hydraulic test machine with different values of support span distances (L = 30, 45, and 60 mm) under various temperature values (T = 23, 40 and 60 °C) in order to analyze the influences of support span and temperature values. The failure mode of the collapsed sandwiches has been investigated using 3D computed tomography (CT) that allows a three-dimensional reconstruction of the analyzed object. The main results of the bending tests are: load-deflection curves, peak force and absorbed energy values. The results were compared according to the effect of cell size, support span and temperature values. The obtained results have particular importance for applications that require lightweight structures with a high capacity of energy dissipation, such as the transport industry, where problems of collision and crash have increased in the last years.

Keywords: light-weight sandwich structures, micro lattice cores, selective laser sintering, transport application

Procedia PDF Downloads 316

Authors: Iryna P. Dzieciuch, Michael D. Putman

Abstract:

Humidity is known to degrade Navy ship electronic equipment, especially in hot moist environments. If left untreated, it can cause significant and permanent damage. Even rigorous inspection and frequent clean-up would not prevent further equipment contamination and degradation because of the constant presence of favorable growth conditions for many microorganisms. Generally, relative humidity levels of less than 60% will inhibit corrosion in electronic equipment, but because NAVY electronics often operate in hot and humid environments, prevention via dehumidification is not always possible. Currently, there is no defined research that fully describes key mechanisms which cause electronics and its coating degradation. The corrosive action of most bacteria is mainly developed through (i) mycelium adherence to the metal plates, (ii) facilitation the formation of pitting areas, (iii) production of organic acids such as citric, iso-citric, cis-aconitic, alpha-ketoglutaric, which are corrosive to electronic equipment and its components. Our approach studies corrosive action in electronic equipment: circuit-board, wires and connections that are exposed in the humid environment that gets worse during condensation. In our new approach the technical task is built on work with the bacterial communities in public areas, bacterial genetics, bioinformatics, biostatistics and Scanning Electron Microscopy (SEM) of corroded circuit boards. Based on these methods, we collect and examine environmental samples from biofilms of the corroded and non-corroded sites, where bacterial contamination of electronic equipment, such as machine racks and shore boats, is an ongoing concern. Sample collection and sample analysis is focused on addressing the key questions identified above through the following tasks: laboratory sample processing and evaluation under scanning electron microscopy, initial sequencing and data evaluation; bioinformatics and data analysis. Preliminary results from scanning electron microscopy (SEM) have revealed that metal particulates and alloys in corroded samples consists mostly of Tin ( < 40%), Silicon ( < 4%), Sulfur ( < 1%), Aluminum ( < 2%), Magnesium ( < 2%), Copper ( < 1%), Bromine ( < 2%), Barium ( <1%) and Iron ( < 2%) elements. We have also performed X 12000 magnification of the same sites and that proved existence of undisrupted biofilm organelles and crystal structures. Non-corrosion sites have revealed high presence of copper ( < 47%); other metals remain at the comparable level as on the samples with corrosion. We have performed X 1000 magnification on the non-corroded at the sites and have documented formation of copper crystals. The next step of this study, is to perform metagenomics sequencing at all sites and to compare bacterial composition present in the environment. While copper is nontoxic to the living organisms, the process of bacterial adhesion creates acidic environment by releasing citric, iso-citric, cis-aconitic, alpha-ketoglutaric acidics, which in turn release copper ions Cu++, which that are highly toxic to the bacteria and higher order living organisms. This phenomenon, might explain natural “antibiotic” properties that are lacking in elements such as tin. To prove or deny this hypothesis we will use next - generation sequencing (NGS) methods to investigate types and growth cycles of bacteria that from bacterial biofilm the on corrosive and non-corrosive samples.

Keywords: bacteria, biofilm, circuit board, copper, corrosion, electronic equipment, organic acids, tin

Procedia PDF Downloads 135

Authors: Qianwei Cheng, A. K. M. Mahbubur Rahman, Anis Sarker, Abu Bakar Siddik Nayem, Ovi Paul, Amin Ahsan Ali, M. Ashraful Amin, Ryosuke Shibasaki, Moinul Zaber

Abstract:

Thomas Friedman, in his famous book, argued that the world in this 21st century is flat and will continue to be flatter. This is attributed to rapid globalization and the interdependence of humanity that engendered tremendous in-flow of human migration towards the urban spaces. In order to keep the urban environment sustainable, policy makers need to plan based on extensive analysis of the urban environment. With the advent of high definition satellite images, high resolution data, computational methods such as deep neural network analysis, and hardware capable of high-speed analysis; urban planning is seeing a paradigm shift. Legacy data on urban environments are now being complemented with high-volume, high-frequency data. However, the first step of understanding urban space lies in useful categorization of the space that is usable for data collection, analysis, and visualization. In this paper, we propose a pragmatic categorization method that is readily usable for machine analysis and show applicability of the methodology on a developing world setting. Categorization to plan sustainable urban spaces should encompass the buildings and their surroundings. However, the state-of-the-art is mostly dominated by classification of building structures, building types, etc. and largely represents the developed world. Hence, these methods and models are not sufficient for developing countries such as Bangladesh, where the surrounding environment is crucial for the categorization. Moreover, these categorizations propose small-scale classifications, which give limited information, have poor scalability and are slow to compute in real time. Our proposed method is divided into two steps-categorization and automation. We categorize the urban area in terms of informal and formal spaces and take the surrounding environment into account. 50 km × 50 km Google Earth image of Dhaka, Bangladesh was visually annotated and categorized by an expert and consequently a map was drawn. The categorization is based broadly on two dimensions-the state of urbanization and the architectural form of urban environment. Consequently, the urban space is divided into four categories: 1) highly informal area; 2) moderately informal area; 3) moderately formal area; and 4) highly formal area. In total, sixteen sub-categories were identified. For semantic segmentation and automatic categorization, Google’s DeeplabV3+ model was used. The model uses Atrous convolution operation to analyze different layers of texture and shape. This allows us to enlarge the field of view of the filters to incorporate larger context. Image encompassing 70% of the urban space was used to train the model, and the remaining 30% was used for testing and validation. The model is able to segment with 75% accuracy and 60% Mean Intersection over Union (mIoU). In this paper, we propose a pragmatic categorization method that is readily applicable for automatic use in both developing and developed world context. The method can be augmented for real-time socio-economic comparative analysis among cities. It can be an essential tool for the policy makers to plan future sustainable urban spaces.

Keywords: semantic segmentation, urban environment, deep learning, urban building, classification

Procedia PDF Downloads 154

Authors: Alexander Vandesompele, Joni Dambre

Abstract:

The diversity in dendritic arborization, as first illustrated by Santiago Ramon y Cajal, has always suggested a role for dendrites in the functionality of neurons. In the past decades, thanks to new recording techniques and optical stimulation methods, it has become clear that dendrites are not merely passive electrical components. They are observed to integrate inputs in a non-linear fashion and actively participate in computations. Regardless, in simulations of neural networks dendritic structure and functionality are often overlooked. Especially in a machine learning context, when designing artificial neural networks, point neuron models such as the leaky-integrate-and-fire (LIF) model are dominant. These models mimic the integration of inputs at the neuron soma, and ignore the existence of dendrites. In this work, the LIF point neuron model is extended with a simple form of dendritic computation. This gives the LIF neuron increased capacity to discriminate spatiotemporal input sequences, a dendritic functionality as observed in another study. Simulations of the spiking neurons are performed using the Bindsnet framework. In the common LIF model, incoming synapses are independent. Here, we introduce a dependency between incoming synapses such that the post-synaptic impact of a spike is not only determined by the weight of the synapse, but also by the activity of other synapses. This is a form of short term plasticity where synapses are potentiated or depressed by the preceding activity of neighbouring synapses. This is a straightforward way to prevent inputs from simply summing linearly at the soma. To implement this, each pair of synapses on a neuron is assigned a variable,representing the synaptic relation. This variable determines the magnitude ofthe short term plasticity. These variables can be chosen randomly or, more interestingly, can be learned using a form of Hebbian learning. We use Spike-Time-Dependent-Plasticity (STDP), commonly used to learn synaptic strength magnitudes. If all neurons in a layer receive the same input, they tend to learn the same through STDP. Adding inhibitory connections between the neurons creates a winner-take-all (WTA) network. This causes the different neurons to learn different input sequences. To illustrate the impact of the proposed dendritic mechanism, even without learning, we attach five input neurons to two output neurons. One output neuron isa regular LIF neuron, the other output neuron is a LIF neuron with dendritic relationships. Then, the five input neurons are allowed to fire in a particular order. The membrane potentials are reset and subsequently the five input neurons are fired in the reversed order. As the regular LIF neuron linearly integrates its inputs at the soma, the membrane potential response to both sequences is similar in magnitude. In the other output neuron, due to the dendritic mechanism, the membrane potential response is different for both sequences. Hence, the dendritic mechanism improves the neuron’s capacity for discriminating spa-tiotemporal sequences. Dendritic computations improve LIF neurons even if the relationships between synapses are established randomly. Ideally however, a learning rule is used to improve the dendritic relationships based on input data. It is possible to learn synaptic strength with STDP, to make a neuron more sensitive to its input. Similarly, it is possible to learn dendritic relationships with STDP, to make the neuron more sensitive to spatiotemporal input sequences. Feeding structured data to a WTA network with dendritic computation leads to a significantly higher number of discriminated input patterns. Without the dendritic computation, output neurons are less specific and may, for instance, be activated by a sequence in reverse order.

Keywords: dendritic computation, spiking neural networks, point neuron model

Procedia PDF Downloads 108

Authors: Martyna Ekiert, Natalia Grzechnik, Joanna Karbowniczek, Urszula Stachewicz, Andrzej Mlyniec

Abstract:

Tendon grafting is a common procedure performed to treat tendon rupture. Before the surgical procedure, tissues intended for grafts (i.e., Achilles tendon) are stored in ultra-low temperatures for a long time and also may be subjected to unfavorable conditions, such as repetitive freezing (F) and thawing (T). Such storage protocols may highly influence the graft mechanical properties, decrease its functionality and thus increase the risk of complications during the transplant procedure. The literature reports on the influence of multiple F/T cycles on internal structure and mechanical properties of tendons stay inconclusive, confirming and denying the negative influence of multiple F/T at the same time. An inconsistent research methodology and lack of clear limit of F/T cycles, which disqualifies tissue for surgical graft purposes, encouraged us to investigate the issue of multiple F/T cycles by the mean of biomechanical tensile tests supported with Scanning Electron Microscope (SEM) imaging. The study was conducted on male bovine Achilles tendon-derived from the local abattoir. Fresh tendons were cleaned of excessive membranes and then sectioned to obtained fascicle bundles. Collected samples were randomly assigned to 6 groups subjected to 1, 2, 4, 6, 8 and 12 cycles of freezing-thawing (F/T), respectively. Each F/T cycle included deep freezing at -80°C temperature, followed by thawing at room temperature. After final thawing, thin slices of the side part of samples subjected to 1, 4, 8 and 12 F/T cycles were collected for SEM imaging. Then, the width and thickness of all samples were measured to calculate the cross-sectional area. Biomechanical tests were performed using the universal testing machine (model Instron 8872, INSTRON®, Norwood, Massachusetts, USA) using a load cell with a maximum capacity of 250 kN and standard atmospheric conditions. Both ends of each fascicle bundle were manually clamped in grasping clamps using abrasive paper and wet cellulose wadding swabs to prevent tissue slipping while clamping and testing. Samples were subjected to the testing procedure including pre-loading, pre-cycling, loading, holding and unloading steps to obtain stress-strain curves for representing tendon stretching and relaxation. The stiffness of AT fascicles bundle samples was evaluated in terms of modulus of elasticity (Young’s modulus), calculated from the slope of the linear region of stress-strain curves. SEM imaging was preceded by chemical sample preparation including 24hr fixation in 3% glutaraldehyde buffered with 0.1 M phosphate buffer, washing with 0.1 M phosphate buffer solution and dehydration in a graded ethanol solution. SEM images (Merlin Gemini II microscope, ZEISS®) were taken using 30 000x mag, which allowed measuring a diameter of collagen fibrils. The results confirm a decrease in fascicle bundles Young’s modulus as well as a decrease in the diameter of collagen fibrils. These results confirm the negative influence of multiple F/T cycles on the mechanical properties of tendon tissue.

Keywords: biomechanics, collagen, fascicle bundles, soft tissue

Procedia PDF Downloads 106

Authors: Marina Passero, Tianhua Zhai, Zuyi (Jacky) Huang

Abstract:

Alzheimer’s disease has become a major public health issue, as indicated by the increasing populations of Americans living with Alzheimer’s disease. After decades of extensive research in Alzheimer’s disease, only seven drugs have been approved by Food and Drug Administration (FDA) to treat Alzheimer’s disease. Five of these drugs were designed to treat the dementia symptoms, and only two drugs (i.e., Aducanumab and Lecanemab) target the progression of Alzheimer’s disease, especially the accumulation of amyloid-b plaques. However, controversial comments were raised for the accelerated approvals of either Aducanumab or Lecanemab, especially with concerns on safety and side effects of these two drugs. There is still an urgent need for further drug discovery to target the biological processes involved in the progression of Alzheimer’s disease. Excessive cholesterol has been found to accumulate in the brain of those with Alzheimer’s disease. Cholesterol can be synthesized in both the blood and the brain, but the majority of biosynthesis in the adult brain takes place in astrocytes and is then transported to the neurons via ApoE. The blood brain barrier separates cholesterol metabolism in the brain from the rest of the body. Various proteins contribute to the metabolism of cholesterol in the brain, which offer potential targets for Alzheimer’s treatment. In the astrocytes, SREBP cleavage-activating protein (SCAP) binds to Sterol Regulatory Element-binding Protein 2 (SREBP2) in order to transport the complex from the endoplasmic reticulum to the Golgi apparatus. Cholesterol is secreted out of the astrocytes by ATP-Binding Cassette A1 (ABCA1) transporter. Lipoprotein receptors such as triggering receptor expressed on myeloid cells 2 (TREM2) internalize cholesterol into the microglia, while lipoprotein receptors such as Low-density lipoprotein receptor-related protein 1 (LRP1) internalize cholesterol into the neuron. Cytochrome P450 Family 46 Subfamily A Member 1 (CYP46A1) converts excess cholesterol to 24S-hydroxycholesterol (24S-OHC). Cholesterol has been approved for its direct effect on the production of amyloid-beta and tau proteins. The addition of cholesterol to the brain promotes the activity of beta-site amyloid precursor protein cleaving enzyme 1 (BACE1), secretase, and amyloid precursor protein (APP), which all aid in amyloid-beta production. The reduction of cholesterol esters in the brain have been found to reduce phosphorylated tau levels in mice. In this work, a computational pipeline was developed to identify the protein targets involved in cholesterol regulation in brain and further to identify chemical compounds as the inhibitors of a selected protein target. Since extensive evidence shows the strong correlation between brain cholesterol regulation and Alzheimer’s disease, a detailed literature review on genes or pathways related to the brain cholesterol synthesis and regulation was first conducted in this work. An interaction network was then built for those genes so that the top gene targets were identified. The involvement of these genes in Alzheimer’s disease progression was discussed, which was followed by the investigation of existing clinical trials for those targets. A ligand-protein docking program was finally developed to screen 1.5 million chemical compounds for the selected protein target. A machine learning program was developed to evaluate and predict the binding interaction between chemical compounds and the protein target. The results from this work pave the way for further drug discovery to regulate brain cholesterol to combat Alzheimer’s disease.

Keywords: Alzheimer’s disease, drug discovery, ligand-protein docking, gene-network analysis, cholesterol regulation

Procedia PDF Downloads 46

Authors: Komal K., Cleary F., Wells J S.G., Bennett L

Abstract:

Smart fabric garments enable various monitoring applications across sectors such as healthcare, sports and fitness, and the military. Healthcare smart garments monitoring EEG, EMG, and ECG rely on the use of electrodes (dry or wet). However, such electrodes, when used for long-term monitoring, can cause discomfort and skin irritation for the wearer because of their inflexible structure and weight. Ongoing research has been investigating textile-based electrodes (textrodes) in order to provide more comfortable and usable fabric-based electrodes capable of providing intuitive biopotential monitoring. Progress has been made in this space, but they still face a critical design challenge in maintaining consistent skin contact, which directly impacts signal quality. Furthermore, there is a lack of an ISO-based testing framework to validate the electrode design and assess its ability to achieve enhanced performance, strength, usability, and durability. This study proposes the development and evaluation of an ISO-compliant testing framework for standard 2D and advanced 3D embroidered textrodes designs that have a unique structure in order to establish enhanced skin contact for the wearer. This testing framework leverages ISO standards: ISO 13934-1:2013 for tensile and zone-wise strength tests; ISO 13937-2 for tear tests; and ISO 6330 for washing, validating the textrode's performance, a necessity for wearables health parameter monitoring applications. Five textrodes (C1-C5) were designed using EPC win digitization software. Varying patterns such as running stitches, lock stitches, back-to-back stitches, and moss stitches were used to create various embroidered tetrodes samples using Madeira HC12 conductive thread with a resistivity of 100 ohm/m. The textrode designs were then fabricated using a ZSK technical embroidery machine. A comparative analysis was conducted based on a series of laboratory tests adhering to ISO compliance requirements. Tests focusing on the application of strain were applied to the textrodes, and these included: (1) analysis of the electrode's overall surface area strength; (2) assessment of the robustness of the textrodes boundaries; and (3) the assignment of fault test zones to each textrode, where vertical and horizontal slits of 3mm were applied to evaluate the performance of textrodes and its durability. Specific ISO-compliant tests linked to washing were conducted multiple times on each textrode sample to assess both mechanical and chemical damage. Additionally, abrasion and pilling tests were performed to evaluate mechanical damage on the surface of the textrodes and to compare it with the washing test. Finally, the textrodes were assessed based on morphological and surface resistance changes. Results demonstrate that textrode C4, featuring a 3-D layered structure consisting of foam, fabric, and conductive thread layers, significantly enhances skin-electrode contact for biopotential recording. The inclusion of a 3D foam layer was particularly effective in maintaining the shape of the electrode during strain tests, making it the top-performing textrode sample. Therefore, the layered 3D design structure of textrode C4 ranks highest when tested for durability, reusability, and washability. The ISO testing framework established in this study will support future research, validating the durability and reliability of textrodes for a wide range of applications.

Keywords: smart fabric, textrodes, testing framework, ISO compliant

Procedia PDF Downloads 40

Authors: A. Peyvan, D. Li, J. Komperda, F. Mashayek

Abstract:

Turbulence can be observed in a variety fluid motions in nature and industrial applications. Recent investment in high-speed aircraft and propulsion systems has revitalized fundamental research on turbulent flows. In these systems, capturing chaotic fluid structures with different length and time scales is accomplished through the Direct Numerical Simulation (DNS) approach since it accurately simulates flows down to smallest dissipative scales, i.e., Kolmogorov’s scales. The discontinuous spectral element method (DSEM) is a high-order technique that uses spectral functions for approximating the solution. The DSEM code has been developed by our research group over the course of more than two decades. Recently, the code has been improved to run large cases in the order of billions of solution points. Running big simulations requires a considerable amount of RAM. Therefore, the DSEM code must be highly parallelized and able to start on multiple computational nodes on an HPC cluster with distributed memory. However, some pre-processing procedures, such as determining global element information, creating a global face list, and assigning global partitioning and element connection information of the domain for communication, must be done sequentially with a single processing core. A separate code has been written to perform the pre-processing procedures on a local machine. It stores the minimum amount of information that is required for the DSEM code to start in parallel, extracted from the mesh file, into text files (pre-files). It packs integer type information with a Stream Binary format in pre-files that are portable between machines. The files are generated to ensure fast read performance on different file-systems, such as Lustre and General Parallel File System (GPFS). A new subroutine has been added to the DSEM code to read the startup files using parallel MPI I/O, for Lustre, in a way that each MPI rank acquires its information from the file in parallel. In case of GPFS, in each computational node, a single MPI rank reads data from the file, which is specifically generated for the computational node, and send them to other ranks on the node using point to point non-blocking MPI communication. This way, communication takes place locally on each node and signals do not cross the switches of the cluster. The read subroutine has been tested on Argonne National Laboratory’s Mira (GPFS), National Center for Supercomputing Application’s Blue Waters (Lustre), San Diego Supercomputer Center’s Comet (Lustre), and UIC’s Extreme (Lustre). The tests showed that one file per node is suited for GPFS and parallel MPI I/O is the best choice for Lustre file system. The DSEM code relies on heavily optimized linear algebra operation such as matrix-matrix and matrix-vector products for calculation of the solution in every time-step. For this, the code can either make use of its matrix math library, BLAS, Intel MKL, or ATLAS. This fact and the discontinuous nature of the method makes the DSEM code run efficiently in parallel. The results of weak scaling tests performed on Blue Waters showed a scalable and efficient performance of the code in parallel computing.

Keywords: computational fluid dynamics, direct numerical simulation, spectral element, turbulent flow

Procedia PDF Downloads 111

Authors: Ajit Rai, Hamza Deroui, Blandine Vacher, Khwansiri Ninpan, Arthur Aumont, Francesco Vitillo, Robert Plana

Abstract:

Large-scale critical industrial scheduling problems are based on Resource-Constrained Project Scheduling Problems (RCPSP), that necessitate integration with other optimization problems (e.g., vehicle routing, supply chain, or unique industrial ones), thus requiring practical solutions (i.e., modular, computationally efficient with feasible solutions). To the best of our knowledge, the current industrial state of the art is not addressing this holistic problem. We propose an original modular solution that answers the issues exhibited by the delivery of complex projects. With three interlinked entities (project, task, resources) having their constraints, it uses a greedy heuristic with a dynamic cost function for each task with a situational assessment at each time step. It handles large-scale data and can be easily integrated with other optimization problems, already existing industrial tools and unique constraints as required by the use case. The solution has been tested and validated by domain experts on three use cases: outage management in Nuclear Power Plants (NPPs), planning of future NPP maintenance operation, and application in the defense industry on supply chain and factory relocation. In the first use case, the solution, in addition to the resources’ availability and tasks’ logical relationships, also integrates several project-specific constraints for outage management, like, handling of resource incompatibility, updating of tasks priorities, pausing tasks in a specific circumstance, and adjusting dynamic unit of resources. With more than 20,000 tasks and multiple constraints, the solution provides a feasible schedule within 10-15 minutes on a standard computer device. This time-effective simulation corresponds with the nature of the problem and requirements of several scenarios (30-40 simulations) before finalizing the schedules. The second use case is a factory relocation project where production lines must be moved to a new site while ensuring the continuity of their production. This generates the challenge of merging job shop scheduling and the RCPSP with location constraints. Our solution allows the automation of the production tasks while considering the rate expectation. The simulation algorithm manages the use and movement of resources and products to respect a given relocation scenario. The last use case establishes a future maintenance operation in an NPP. The project contains complex and hard constraints, like on Finish-Start precedence relationship (i.e., successor tasks have to start immediately after predecessors while respecting all constraints), shareable coactivity for managing workspaces, and requirements of a specific state of "cyclic" resources (they can have multiple states possible with only one at a time) to perform tasks (can require unique combinations of several cyclic resources). Our solution satisfies the requirement of minimization of the state changes of cyclic resources coupled with the makespan minimization. It offers a solution of 80 cyclic resources with 50 incompatibilities between levels in less than a minute. Conclusively, we propose a fast and feasible modular approach to various industrial scheduling problems that were validated by domain experts and compatible with existing industrial tools. This approach can be further enhanced by the use of machine learning techniques on historically repeated tasks to gain further insights for delay risk mitigation measures.

Keywords: deterministic scheduling, optimization coupling, modular scheduling, RCPSP

Procedia PDF Downloads 165

Authors: Utkarsh A. Mishra, Ankit Bansal

Abstract:

At high temperature, radiative heat transfer is the dominant mode of heat transfer. It is governed by various phenomena such as photon emission, absorption, and scattering. The solution of the governing integrodifferential equation of radiative transfer is a complex process, more when the effect of participating medium and wavelength properties are taken into consideration. Although a generic formulation of such radiative transport problem can be modeled for a wide variety of problems with non-gray, non-diffusive surfaces, there is always a trade-off between simplicity and accuracy of the problem. Recently, solutions of complicated mathematical problems with statistical methods based on randomization of naturally occurring phenomena have gained significant importance. Photon bundles with discrete energy can be replicated with random numbers describing the emission, absorption, and scattering processes. Photon Monte Carlo (PMC) is a simple, yet powerful technique, to solve radiative transfer problems in complicated geometries with arbitrary participating medium. The method, on the one hand, increases the accuracy of estimation, and on the other hand, increases the computational cost. The participating media -generally a gas, such as CO₂, CO, and H₂O- present complex emission and absorption spectra. To model the emission/absorption accurately with random numbers requires a weighted sampling as different sections of the spectrum carries different importance. Importance sampling (IS) was implemented to sample random photon of arbitrary wavelength, and the sampled data provided unbiased training of MC estimators for better results. A better replacement to uniform random numbers is using deterministic, quasi-random sequences. Halton, Sobol, and Faure Low-Discrepancy Sequences are used in this study. They possess better space-filling performance than the uniform random number generator and gives rise to a low variance, stable Quasi-Monte Carlo (QMC) estimators with faster convergence. An optimal supervised learning scheme was further considered to reduce the computation costs of the PMC simulation. A one-dimensional plane-parallel slab problem with participating media was formulated. The history of some randomly sampled photon bundles is recorded to train an Artificial Neural Network (ANN), back-propagation model. The flux was calculated using the standard quasi PMC and was considered to be the training target. Results obtained with the proposed model for the one-dimensional problem are compared with the exact analytical and PMC model with the Line by Line (LBL) spectral model. The approximate variance obtained was around 3.14%. Results were analyzed with respect to time and the total flux in both cases. A significant reduction in variance as well a faster rate of convergence was observed in the case of the QMC method over the standard PMC method. However, the results obtained with the ANN method resulted in greater variance (around 25-28%) as compared to the other cases. There is a great scope of machine learning models to help in further reduction of computation cost once trained successfully. Multiple ways of selecting the input data as well as various architectures will be tried such that the concerned environment can be fully addressed to the ANN model. Better results can be achieved in this unexplored domain.

Keywords: radiative heat transfer, Monte Carlo Method, pseudo-random numbers, low discrepancy sequences, artificial neural networks

Procedia PDF Downloads 199

Authors: Claudio Loconsole, Daniele Leonardis, Antonio Brunetti, Gianpaolo Francesco Trotta, Nicholas Caporusso, Vitoantonio Bevilacqua

Abstract:

Unfortunately, many people still do not have access to communication, with specific regard to reading and writing. Among them, people who are blind or visually impaired, have several difficulties in getting access to the world, compared to the sighted. Indeed, despite technology advancement and cost reduction, nowadays assistive devices are still expensive such as Braille-based input/output systems which enable reading and writing texts (e.g., personal notes, documents). As a consequence, assistive technology affordability is fundamental in supporting the visually impaired in communication, learning, and social inclusion. This, in turn, has serious consequences in terms of equal access to opportunities, freedom of expression, and actual and independent participation to a society designed for the sighted. Moreover, the visually impaired experience difficulties in recognizing objects and interacting with devices in any activities of daily living. It is not a case that Braille indications are commonly reported only on medicine boxes and elevator keypads. Several software applications for the automatic translation of written text into speech (e.g., Text-To-Speech - TTS) enable reading pieces of documents. However, apart from simple tasks, in many circumstances TTS software is not suitable for understanding very complicated pieces of text requiring to dwell more on specific portions (e.g., mathematical formulas or Greek text). In addition, the experience of reading\writing text is completely different both in terms of engagement, and from an educational perspective. Statistics on the employment rate of blind people show that learning to read and write provides the visually impaired with up to 80% more opportunities of finding a job. Especially in higher educational levels, where the ability to digest very complex text is key, accessibility and availability of Braille plays a fundamental role in reducing drop-out rate of the visually impaired, thus affecting the effectiveness of the constitutional right to get access to education. In this context, the Braille Lab project aims at overcoming these social needs by including affordability in designing and developing assistive tools for visually impaired people. In detail, our awarded project focuses on a technology innovation of the operation principle of existing assistive tools for the visually impaired leaving the Human-Machine Interface unchanged. This can result in a significant reduction of the production costs and consequently of tool selling prices, thus representing an important opportunity for social entrepreneurship. The first two assistive tools designed within the Braille Lab project following the proposed approach aims to provide the possibility to personally print documents and handouts and to read texts written in Braille using refreshable Braille display, respectively. The former, named ‘Braille Cartridge’, represents an alternative solution for printing in Braille and consists in the realization of an electronic-controlled dispenser printing (cartridge) which can be integrated within traditional ink-jet printers, in order to leverage the efficiency and cost of the device mechanical structure which are already being used. The latter, named ‘Braille Cursor’, is an innovative Braille display featuring a substantial technology innovation by means of a unique cursor virtualizing Braille cells, thus limiting the number of active pins needed for Braille characters.

Keywords: Human rights, social challenges and technology innovations, visually impaired, affordability, assistive tools

Procedia PDF Downloads 246

Authors: Hyohun Kim, Dongwha Shin, Yeonseok Kim, Ji-Su Ahn, Kensuke Nakamura, Dongeun Choi, Byung-Woo Hong

Abstract:

Artefacts are commonly encountered in the imaging process of clinical computed tomography (CT) where the artefact refers to any systematic discrepancy between the reconstructed observation and the true attenuation coefficient of the object. It is known that CT images are inherently more prone to artefacts due to its image formation process where a large number of independent detectors are involved, and they are assumed to yield consistent measurements. There are a number of different artefact types including noise, beam hardening, scatter, pseudo-enhancement, motion, helical, ring, and metal artefacts, which cause serious difficulties in reading images. Thus, it is desired to remove nuisance factors from the degraded image leaving the fundamental intrinsic information that can provide better interpretation of the anatomical and pathological characteristics. However, it is considered as a difficult task due to the high dimensionality and variability of data to be recovered, which naturally motivates the use of machine learning techniques. We propose an image restoration algorithm based on the deep neural network framework where the denoising auto-encoders are stacked building multiple layers. The denoising auto-encoder is a variant of a classical auto-encoder that takes an input data and maps it to a hidden representation through a deterministic mapping using a non-linear activation function. The latent representation is then mapped back into a reconstruction the size of which is the same as the size of the input data. The reconstruction error can be measured by the traditional squared error assuming the residual follows a normal distribution. In addition to the designed loss function, an effective regularization scheme using residual-driven dropout determined based on the gradient at each layer. The optimal weights are computed by the classical stochastic gradient descent algorithm combined with the back-propagation algorithm. In our algorithm, we initially decompose an input image into its intrinsic representation and the nuisance factors including artefacts based on the classical Total Variation problem that can be efficiently optimized by the convex optimization algorithm such as primal-dual method. The intrinsic forms of the input images are provided to the deep denosing auto-encoders with their original forms in the training phase. In the testing phase, a given image is first decomposed into the intrinsic form and then provided to the trained network to obtain its reconstruction. We apply our algorithm to the restoration of the corrupted CT images by the artefacts. It is shown that our algorithm improves the readability and enhances the anatomical and pathological properties of the object. The quantitative evaluation is performed in terms of the PSNR, and the qualitative evaluation provides significant improvement in reading images despite degrading artefacts. The experimental results indicate the potential of our algorithm as a prior solution to the image interpretation tasks in a variety of medical imaging applications. This work was supported by the MISP(Ministry of Science and ICT), Korea, under the National Program for Excellence in SW (20170001000011001) supervised by the IITP(Institute for Information and Communications Technology Promotion).

Keywords: auto-encoder neural network, CT image artefact, deep learning, intrinsic image representation, noise reduction, total variation

Procedia PDF Downloads 169

Authors: Joseph M. Rich, Vinay A. Duddalwar, Assad A. Oberai

Abstract:

Prostate adenocarcinoma is the most common cancer in males, with osseous metastases as the commonest site of metastatic prostate carcinoma (mPC). Treatment monitoring is based on the evaluation and characterization of lesions on multiple imaging studies, including Computed Tomography (CT). Monitoring of the osseous disease burden, including follow-up of lesions and identification and characterization of new lesions, is a laborious task for radiologists. Deep learning algorithms are increasingly used to perform tasks such as identification and segmentation for osseous metastatic disease and provide accurate information regarding metastatic burden. Here, nnUNet was used to produce a model which can segment CT scan images of prostate adenocarcinoma vertebral bone metastatic lesions. nnUNet is an open-source Python package that adds optimizations to deep learning-based UNet architecture but has not been extensively combined with transfer learning techniques due to the absence of a readily available functionality of this method. The IRB-approved study data set includes imaging studies from patients with mPC who were enrolled in clinical trials at the University of Southern California (USC) Health Science Campus and Los Angeles County (LAC)/USC medical center. Manual segmentation of metastatic lesions was completed by an expert radiologist Dr. Vinay Duddalwar (20+ years in radiology and oncologic imaging), to serve as ground truths for the automated segmentation. Despite nnUNet’s success on some medical segmentation tasks, it only produced an average Dice Similarity Coefficient (DSC) of 0.31 on the USC dataset. DSC results fell in a bimodal distribution, with most scores falling either over 0.66 (reasonably accurate) or at 0 (no lesion detected). Applying more aggressive data augmentation techniques dropped the DSC to 0.15, and reducing the number of epochs reduced the DSC to below 0.1. Datasets have been identified for transfer learning, which involve balancing between size and similarity of the dataset. Identified datasets include the Pancreas data from the Medical Segmentation Decathlon, Pelvic Reference Data, and CT volumes with multiple organ segmentations (CT-ORG). Some of the challenges of producing an accurate model from the USC dataset include small dataset size (115 images), 2D data (as nnUNet generally performs better on 3D data), and the limited amount of public data capturing annotated CT images of bone lesions. Optimizations and improvements will be made by applying transfer learning and generative methods, including incorporating generative adversarial networks and diffusion models in order to augment the dataset. Performance with different libraries, including MONAI and custom architectures with Pytorch, will be compared. In the future, molecular correlations will be tracked with radiologic features for the purpose of multimodal composite biomarker identification. Once validated, these models will be incorporated into evaluation workflows to optimize radiologist evaluation. Our work demonstrates the challenges of applying automated image segmentation to small medical datasets and lays a foundation for techniques to improve performance. As machine learning models become increasingly incorporated into the workflow of radiologists, these findings will help improve the speed and accuracy of vertebral metastatic lesions detection.

Keywords: deep learning, image segmentation, medicine, nnUNet, prostate carcinoma, radiomics

Procedia PDF Downloads 73

Authors: Yunwei Tang, Linhai Jing, Hui Li, Qingjie Liu, Xiuxia Li, Qi Yan, Haifeng Ding

Abstract:

The giant panda (Ailuropoda melanoleuca) is an endangered species, mainly live in central China, where bamboos act as the main food source of wild giant pandas. Knowledge of spatial distribution of bamboos therefore becomes important for identifying the habitat of giant pandas. There have been ongoing studies for mapping bamboos and other tree species using remote sensing. WorldView-2 (WV-2) is the first high resolution commercial satellite with eight Multi-Spectral (MS) bands. Recent studies demonstrated that WV-2 imagery has a high potential in classification of tree species. The advanced classification techniques are important for utilising high spatial resolution imagery. It is generally agreed that object-based image analysis is a more desirable method than pixel-based analysis in processing high spatial resolution remotely sensed data. Classifiers that use spatial information combined with spectral information are known as contextual classifiers. It is suggested that contextual classifiers can achieve greater accuracy than non-contextual classifiers. Thus, spatial correlation can be incorporated into classifiers to improve classification results. The study area is located at Wuyipeng area in Wolong, Sichuan Province. The complex environment makes it difficult for information extraction since bamboos are sparsely distributed, mixed with brushes, and covered by other trees. Extensive fieldworks in Wuyingpeng were carried out twice. The first one was on 11th June, 2014, aiming at sampling feature locations for geometric correction and collecting training samples for classification. The second fieldwork was on 11th September, 2014, for the purposes of testing the classification results. In this study, spectral separability analysis was first performed to select appropriate MS bands for classification. Also, the reflectance analysis provided information for expanding sample points under the circumstance of knowing only a few. Then, a spatially weighted object-based k-nearest neighbour (k-NN) classifier was applied to the selected MS bands to identify seven land cover types (bamboo, conifer, broadleaf, mixed forest, brush, bare land, and shadow), accounting for spatial correlation within classes using geostatistical modelling. The spatially weighted k-NN method was compared with three alternatives: the traditional k-NN classifier, the Support Vector Machine (SVM) method and the Classification and Regression Tree (CART). Through field validation, it was proved that the classification result obtained using the spatially weighted k-NN method has the highest overall classification accuracy (77.61%) and Kappa coefficient (0.729); the producer’s accuracy and user’s accuracy achieve 81.25% and 95.12% for the bamboo class, respectively, also higher than the other methods. Photos of tree crowns were taken at sample locations using a fisheye camera, so the canopy density could be estimated. It is found that it is difficult to identify bamboo in the areas with a large canopy density (over 0.70); it is possible to extract bamboos in the areas with a median canopy density (from 0.2 to 0.7) and in a sparse forest (canopy density is less than 0.2). In summary, this study explores the ability of WV-2 imagery for bamboo extraction in a mountainous region in Sichuan. The study successfully identified the bamboo distribution, providing supporting knowledge for assessing the habitats of giant pandas.

Keywords: bamboo mapping, classification, geostatistics, k-NN, worldview-2

Procedia PDF Downloads 293

Authors: Dongping Fang, Lian Duan, Xiaojing Yuan, Mike Xu, Allyn Klunder, Kevin Tan, Suiting Cao, Yeqing Ji

Abstract:

Accurate prediction of a medical condition with straight clinical evidence is a long-sought topic in the medical management and health insurance field. Although great progress has been made with machine learning algorithms, the medical community is still, to a certain degree, suspicious about the model's accuracy and interpretability. This paper presents an innovative hierarchical attention deep learning model to achieve good prediction and clear interpretability that can be easily understood by medical professionals. This deep learning model uses a hierarchical attention structure that matches naturally with the medical history data structure and reflects the member’s encounter (date of service) sequence. The model attention structure consists of 3 levels: (1) attention on the medical code types (diagnosis codes, procedure codes, lab test results, and prescription drugs), (2) attention on the sequential medical encounters within a type, (3) attention on the medical codes within an encounter and type. This model is applied to predict the occurrence of stage 3 chronic kidney disease (CKD3), using three years’ medical history of Medicare Advantage (MA) members from a top health insurance company. The model takes members’ medical events, both claims and electronic medical record (EMR) data, as input, makes a prediction of CKD3 and calculates the contribution from individual events to the predicted outcome. The model outcome can be easily explained with the clinical evidence identified by the model algorithm. Here are examples: Member A had 36 medical encounters in the past three years: multiple office visits, lab tests and medications. The model predicts member A has a high risk of CKD3 with the following well-contributed clinical events - multiple high ‘Creatinine in Serum or Plasma’ tests and multiple low kidneys functioning ‘Glomerular filtration rate’ tests. Among the abnormal lab tests, more recent results contributed more to the prediction. The model also indicates regular office visits, no abnormal findings of medical examinations, and taking proper medications decreased the CKD3 risk. Member B had 104 medical encounters in the past 3 years and was predicted to have a low risk of CKD3, because the model didn’t identify diagnoses, procedures, or medications related to kidney disease, and many lab test results, including ‘Glomerular filtration rate’ were within the normal range. The model accurately predicts members A and B and provides interpretable clinical evidence that is validated by clinicians. Without extra effort, the interpretation is generated directly from the model and presented together with the occurrence date. Our model uses the medical data in its most raw format without any further data aggregation, transformation, or mapping. This greatly simplifies the data preparation process, mitigates the chance for error and eliminates post-modeling work needed for traditional model explanation. To our knowledge, this is the first paper on an interpretable deep-learning model using a 3-level attention structure, sourcing both EMR and claim data, including all 4 types of medical data, on the entire Medicare population of a big insurance company, and more importantly, directly generating model interpretation to support user decision. In the future, we plan to enrich the model input by adding patients’ demographics and information from free-texted physician notes.

Keywords: deep learning, interpretability, attention, big data, medical conditions

Procedia PDF Downloads 74

Authors: Anjalika Fernando, Banuka Athuraliya

Abstract:

This paper presents a comprehensive approach for converting Sinhala Sign Language (SSL) into grammatically correct sentences using Natural Language Processing (NLP) techniques in real-time. While previous studies have explored various aspects of SSL translation, the research gap lies in the absence of grammar checking for SSL. This work aims to bridge this gap by proposing a two-stage methodology that leverages deep learning models to detect signs and translate them into coherent sentences, ensuring grammatical accuracy. The first stage of the approach involves the utilization of a Long Short-Term Memory (LSTM) deep learning model to recognize and interpret SSL signs. By training the LSTM model on a dataset of SSL gestures, it learns to accurately classify and translate these signs into textual representations. The LSTM model achieves a commendable accuracy rate of 94%, demonstrating its effectiveness in accurately recognizing and translating SSL gestures. Building upon the successful recognition and translation of SSL signs, the second stage of the methodology focuses on improving the grammatical correctness of the translated sentences. The project employs a Neural Machine Translation (NMT) architecture, consisting of an encoder and decoder with LSTM components, to enhance the syntactical structure of the generated sentences. By training the NMT model on a parallel corpus of Sinhala wrong sentences and their corresponding grammatically correct translations, it learns to generate coherent and grammatically accurate sentences. The NMT model achieves an impressive accuracy rate of 98%, affirming its capability to produce linguistically sound translations. The proposed approach offers significant contributions to the field of SSL translation and grammar correction. Addressing the critical issue of grammar checking, it enhances the usability and reliability of SSL translation systems, facilitating effective communication between hearing-impaired and non-sign language users. Furthermore, the integration of deep learning techniques, such as LSTM and NMT, ensures the accuracy and robustness of the translation process. This research holds great potential for practical applications, including educational platforms, accessibility tools, and communication aids for the hearing-impaired. Furthermore, it lays the foundation for future advancements in SSL translation systems, fostering inclusive and equal opportunities for the deaf community. Future work includes expanding the existing datasets to further improve the accuracy and generalization of the SSL translation system. Additionally, the development of a dedicated mobile application would enhance the accessibility and convenience of SSL translation on handheld devices. Furthermore, efforts will be made to enhance the current application for educational purposes, enabling individuals to learn and practice SSL more effectively. Another area of future exploration involves enabling two-way communication, allowing seamless interaction between sign-language users and non-sign-language users.In conclusion, this paper presents a novel approach for converting Sinhala Sign Language gestures into grammatically correct sentences using NLP techniques in real time. The two-stage methodology, comprising an LSTM model for sign detection and translation and an NMT model for grammar correction, achieves high accuracy rates of 94% and 98%, respectively. By addressing the lack of grammar checking in existing SSL translation research, this work contributes significantly to the development of more accurate and reliable SSL translation systems, thereby fostering effective communication and inclusivity for the hearing-impaired community

Keywords: Sinhala sign language, sign Language, NLP, LSTM, NMT

Procedia PDF Downloads 76

Authors: Sherif Keddis, Rafik Mitry, Norbert Schwesinger

Abstract:

Intelligent networking of devices has rapidly been gaining importance over the past years and with recent advances in the fields of microcontrollers, integrated circuits and wireless communication, low power applications have emerged, enabling this trend even more. Connected devices provide a much larger database thus enabling highly intelligent and accurate systems. Ensuring safe drinking water is one of the fields that require constant monitoring and can benefit from an increased accuracy. Monitoring is mainly achieved either through complex measures, such as collecting samples from the points of use, or through metering systems typically distant to the points of use which deliver less accurate assessments of the quality of water. Constant metering near the points of use is complicated due to their inaccessibility; e.g. buried water pipes, locked spaces, which makes system maintenance extremely difficult and often unviable. The research presented here attempts to overcome this challenge by providing these systems with enough energy through a flow harvester inside the pipe thus eliminating the maintenance requirements in terms of battery replacements or containment of leakage resulting from wiring such systems. The proposed flow harvester exploits the piezoelectric properties of polyvinylidene difluoride (PVDF) films to convert turbulence induced oscillations into electrical energy. It is intended to be used in standard water pipes with diameters between 0.5 and 1 inch. The working principle of the harvester uses a ring shaped bluff body inside the pipe to induce pressure fluctuations. Additionally the bluff body houses electronic components such as storage, circuitry and RF-unit. Placing the piezoelectric films downstream of that bluff body causes their oscillation which generates electrical charge. The PVDF-film is placed as a multilayered wrap fixed to the pipe wall leaving the top part to oscillate freely inside the flow. The warp, which allows for a larger active, consists of two layers of 30µm thick and 12mm wide PVDF layered alternately with two centered 6µm thick and 8mm wide aluminum foil electrodes. The length of the layers depends on the number of windings and is part of the investigation. Sealing the harvester against liquid penetration is achieved by wrapping it in a ring-shaped LDPE-film and welding the open ends. The fabrication of the PVDF-wraps is done by hand. After validating the working principle using a wind tunnel, experiments have been conducted in water, placing the harvester inside a 1 inch pipe at water velocities of 0.74m/s. To find a suitable placement of the wrap inside the pipe, two forms of fixation were compared regarding their power output. Further investigations regarding the number of windings required for efficient transduction were made. Best results were achieved using a wrap with 3 windings of the active layers which delivers a constant power output of 0.53µW at a 2.3MΩ load and an effective voltage of 1.1V. Considering the extremely low power requirements of sensor applications, these initial results are promising. For further investigations and optimization, machine designs are currently being developed to automate the fabrication and decrease tolerance of the prototypes.

Keywords: maintenance-free sensors, measurements at point of use, piezoelectric flow harvesting, universal micro generator, wireless metering systems

Procedia PDF Downloads 166

Authors: Victoria L. Chester, Usha Kuruganti

Abstract:

The forest industry is a major economic sector of Canada and also one of the most dangerous industries for workers. The use of mechanized mobile forestry harvesting machines has successfully reduced the incidence of injuries in forest workers related to manual labor. However, these machines have also created additional concerns, including a high machine operation learning curve, increased the length of the workday, repetitive strain injury, cognitive load, physical and mental fatigue, and increased postural loads due to sitting in a confined space. It is critical to obtain objective performance data for employers to develop appropriate work practices for this industry, however ergonomic field studies of this industry are lacking mainly due to the difficulties in obtaining comprehensive data while operators are cutting trees in the woods. The purpose of this study was to establish a measurement and experimental protocol to examine the effects of worker skill level and movement training speed (joystick crane speed) on harvesting performance using a forestry simulator. A custom wrist angle measurement device was developed as part of the study to monitor Euler angles during operation of the simulator. The device of the system consisted of two accelerometers, a Bluetooth module, three 3V coin cells, a microcontroller, a voltage regulator and an application software. Harvesting performance and crane data was provided by the simulator software and included tree to frame collisions, crane to tree collisions, boom tip distance, number of trees cut, etc. A pilot study of 3 operators with various skill levels was tested to identify factors that distinguish highly skilled operators from novice or intermediate operators. Dependent variables such as reaction time, math skill, past work experience, training movement speed (e.g. joystick control speeds), harvesting experience level, muscle activity, and wrist biomechanics were measured and analyzed. A 10-channel wireless surface EMG system was used to monitor the amplitude and mean frequency of 10 upper extremity muscles during pre and postperformance on the forestry harvest stimulator. The results of the pilot study showed inconsistent changes in median frequency pre-and postoperation, but there was the increase in the activity of the flexor carpi radialis, anterior deltoid and upper trapezius of both arms. The wrist sensor results indicated that wrist supination and pronation occurred more than flexion and extension with radial-ulnar rotation demonstrating the least movement. Overall, wrist angular motion increased as the crane speed increased from slow to fast. Further data collection is needed and will help industry partners determine those factors that separate skill levels of operators, identify optimal training speeds, and determine the length of training required to bring new operators to an efficient skill level effectively. In addition to effective and employment training programs, results of this work will be used for selective employee recruitment strategies to improve employee retention after training. Further, improved training procedures and knowledge of the physical and mental demands on workers will lead to highly trained and efficient personnel, reduced risk of injury, and optimal work protocols.

Keywords: EMG, forestry, human factors, wrist biomechanics

Procedia PDF Downloads 121

Authors: Zeynep Busra Velioglu, Deniz Pulat, Beril Demirbakan, Burak Ozcan, Ece Bayrak, Cevat Erisken

Abstract:

Bone tissue has the ability to perform a wide array of functions including providing posture, load-bearing capacity, protection for the internal organs, initiating hematopoiesis, and maintaining the homeostasis of key electrolytes via calcium/phosphate ion storage. The most common cause for bone defects is extensive trauma and subsequent infection. Bone tissue has the self-healing capability without a scar tissue formation for the majority of the injuries. However, some may result with delayed union or fracture non-union. Such cases include reconstruction of large bone defects or cases of compromised regenerative process as a result of avascular necrosis and osteoporosis. Several surgical methods exist to treat bone defects, including Ilizarov method, Masquelete technique, growth factor stimulation, and bone replacement. Unfortunately, these are technically demanding and come with noteworthy disadvantages such as lengthy treatment duration, adverse effects on the patient’s psychology, repeated surgical procedures, and often long hospitalization times. These limitations associated with surgical techniques make bone substitutes an attractive alternative. Here, it was hypothesized that a 3D printed scaffold will mimic trabecular bone in terms of biomechanical properties and that such scaffolds will support cell attachment and survival. To test this hypothesis, this study aimed at fabricating poly(lactic acid), PLA, structures using 3D printing technology for trabecular bone defects, characterizing the scaffolds and comparing with bovine trabecular bone. Capacity of scaffolds on human bone marrow stem cell (hBMSC) attachment and survival was also evaluated. Cubes with a volume of 1 cm³ having pore sizes of 0.50, 1.00 and 1.25 mm were printed. The scaffolds/grafts were characterized in terms of porosity, contact angle, compressive mechanical properties as well cell response. Porosities of the 3D printed scaffolds were calculated based on apparent densities. For contact angles, 50 µl distilled water was dropped over the surface of scaffolds, and contact angles were measured using ‘Image J’ software. Mechanical characterization under compression was performed on scaffolds and native trabecular bone (bovine, 15 months) specimens using a universal testing machine at a rate of 0.5mm/min. hBMSCs were seeded onto the 3D printed scaffolds. After 3 days of incubation with fully supplemented Dulbecco’s modified Eagle’s medium, the cells were fixed using 2% formaldehyde and glutaraldehyde mixture. The specimens were then imaged under scanning electron microscopy. Cell proliferation was determined by using EZQuant dsDNA Quantitation kit. Fluorescence was measured using microplate reader Spectramax M2 at the excitation and emission wavelengths of 485nm and 535nm, respectively. Findings suggested that porosity of scaffolds with pore dimensions of 0.5mm, 1.0mm and 1.25mm were not affected by pore size, while contact angle and compressive modulus decreased with increasing pore size. Biomechanical characterization of trabecular bone yielded higher modulus values as compared to scaffolds with all pore sizes studied. Cells attached and survived in all surfaces, demonstrating higher proliferation on scaffolds with 1.25mm pores as compared with those of 1mm. Collectively, given lower mechanical properties of scaffolds as compared to native bone, and biocompatibility of the scaffolds, the 3D printed PLA scaffolds of this study appear as candidate substitutes for bone repair and regeneration.

Keywords: 3D printing, biomechanics, bone repair, stem cell

Procedia PDF Downloads 155

Authors: Denis Jordan, Daniel Golkowski, Mathias Lukas, Katharina Merz, Caroline Mlynarcik, Max Maurer, Valentin Riedl, Stefan Foerster, Eberhard F. Kochs, Andreas Bender, Ruediger Ilg

Abstract:

Introduction: So far, clinical assessments that rely on behavioral responses to differentiate coma states or even predict outcome in coma patients are unreliable, e.g. because of some patients’ motor disabilities. The present study was aimed to provide prognosis in coma patients using markers from electroencephalogram (EEG), blood oxygen level dependent (BOLD) functional magnetic resonance imaging (fMRI) and [18F]-fluorodeoxyglucose (FDG) positron emission tomography (PET). Unsuperwised principal component analysis (PCA) was used for multimodal integration of markers. Methods: Approved by the local ethics committee of the Technical University of Munich (Germany) 20 patients (aged 18-89) with severe brain damage were acquired through intensive care units at the Klinikum rechts der Isar in Munich and at the Therapiezentrum Burgau (Germany). At the day of EEG/fMRI/PET measurement (date I) patients (<3.5 month in coma) were grouped in the minimal conscious state (MCS) or vegetative state (VS) on the basis of their clinical presentation (coma recovery scale-revised, CRS-R). Follow-up assessment (date II) was also based on CRS-R in a period of 8 to 24 month after date I. At date I, 63 channel EEG (Brain Products, Gilching, Germany) was recorded outside the scanner, and subsequently simultaneous FDG-PET/fMRI was acquired on an integrated Siemens Biograph mMR 3T scanner (Siemens Healthineers, Erlangen Germany). Power spectral densities, permutation entropy (PE) and symbolic transfer entropy (STE) were calculated in/between frontal, temporal, parietal and occipital EEG channels. PE and STE are based on symbolic time series analysis and were already introduced as robust markers separating wakefulness from unconsciousness in EEG during general anesthesia. While PE quantifies the regularity structure of the neighboring order of signal values (a surrogate of cortical information processing), STE reflects information transfer between two signals (a surrogate of directed connectivity in cortical networks). fMRI was carried out using SPM12 (Wellcome Trust Center for Neuroimaging, University of London, UK). Functional images were realigned, segmented, normalized and smoothed. PET was acquired for 45 minutes in list-mode. For absolute quantification of brain’s glucose consumption rate in FDG-PET, kinetic modelling was performed with Patlak’s plot method. BOLD signal intensity in fMRI and glucose uptake in PET was calculated in 8 distinct cortical areas. PCA was performed over all markers from EEG/fMRI/PET. Prognosis (persistent VS and deceased patients vs. recovery to MCS/awake from date I to date II) was evaluated using the area under the curve (AUC) including bootstrap confidence intervals (CI, *: p<0.05). Results: Prognosis was reliably indicated by the first component of PCA (AUC=0.99*, CI=0.92-1.00) showing a higher AUC when compared to the best single markers (EEG: AUC<0.96*, fMRI: AUC<0.86*, PET: AUC<0.60). CRS-R did not show prediction (AUC=0.51, CI=0.29-0.78). Conclusion: In a multimodal analysis of EEG/fMRI/PET in coma patients, PCA lead to a reliable prognosis. The impact of this result is evident, as clinical estimates of prognosis are inapt at time and could be supported by quantitative biomarkers from EEG, fMRI and PET. Due to the small sample size, further investigations are required, in particular allowing superwised learning instead of the basic approach of unsuperwised PCA.

Keywords: coma states and prognosis, electroencephalogram, entropy, functional magnetic resonance imaging, machine learning, positron emission tomography, principal component analysis

Procedia PDF Downloads 316

Authors: Qiaosi Li, Frankie Kwan Kit Wong, Tung Fung

Abstract:

Lightweight unmanned aerial vehicle (UAV) loading with novel sensors offers a low cost approach for data acquisition in complex environment. This study established a framework for applying UAV system in complex environment quick mapping and assessed the performance of UAV-based hyperspectral image and digital surface model (DSM) derived from photogrammetric point clouds for 13 species classification in wetland area Mai Po Inner Deep Bay Ramsar Site, Hong Kong. The study area was part of shallow bay with flat terrain and the major species including reedbed and four mangroves: Kandelia obovata, Aegiceras corniculatum, Acrostichum auerum and Acanthus ilicifolius. Other species involved in various graminaceous plants, tarbor, shrub and invasive species Mikania micrantha. In particular, invasive species climbed up to the mangrove canopy caused damage and morphology change which might increase species distinguishing difficulty. Hyperspectral images were acquired by Headwall Nano sensor with spectral range from 400nm to 1000nm and 0.06m spatial resolution image. A sequence of multi-view RGB images was captured with 0.02m spatial resolution and 75% overlap. Hyperspectral image was corrected for radiative and geometric distortion while high resolution RGB images were matched to generate maximum dense point clouds. Furtherly, a 5 cm grid digital surface model (DSM) was derived from dense point clouds. Multiple feature reduction methods were compared to identify the efficient method and to explore the significant spectral bands in distinguishing different species. Examined methods including stepwise discriminant analysis (DA), support vector machine (SVM) and minimum noise fraction (MNF) transformation. Subsequently, spectral subsets composed of the first 20 most importance bands extracted by SVM, DA and MNF, and multi-source subsets adding extra DSM to 20 spectrum bands were served as input in maximum likelihood classifier (MLC) and SVM classifier to compare the classification result. Classification results showed that feature reduction methods from best to worst are MNF transformation, DA and SVM. MNF transformation accuracy was even higher than all bands input result. Selected bands frequently laid along the green peak, red edge and near infrared. Additionally, DA found that chlorophyll absorption red band and yellow band were also important for species classification. In terms of 3D data, DSM enhanced the discriminant capacity among low plants, arbor and mangrove. Meanwhile, DSM largely reduced misclassification due to the shadow effect and morphological variation of inter-species. In respect to classifier, nonparametric SVM outperformed than MLC for high dimension and multi-source data in this study. SVM classifier tended to produce higher overall accuracy and reduce scattered patches although it costs more time than MLC. The best result was obtained by combining MNF components and DSM in SVM classifier. This study offered a precision species distribution survey solution for inaccessible wetland area with low cost of time and labour. In addition, findings relevant to the positive effect of DSM as well as spectral feature identification indicated that the utility of UAV-borne hyperspectral and photogrammetry deriving 3D data is promising in further research on wetland species such as bio-parameters modelling and biological invasion monitoring.

Keywords: digital surface model (DSM), feature reduction, hyperspectral, photogrammetric point cloud, species mapping, unmanned aerial vehicle (UAV)

Procedia PDF Downloads 232

Authors: Surbhi Kaura, Sachin Kandhari, Shahiduz Zafar

Abstract:

Chronic stroke, characterized by persistent motor deficits, often necessitates comprehensive rehabilitation interventions to improve functional outcomes and mitigate long-term dependency. Luna neurorobotic devices, integrated with EMG feedback systems, provide an innovative platform for facilitating neuroplasticity and functional improvement in stroke survivors. This retrospective study aims to investigate the impact of EMG-based Luna neurorobotic interventions on the strengthening of the affected side in chronic stroke patients. In rehabilitation, active patient participation significantly activates the sensorimotor network during motor control, unlike passive movement. Stroke is a debilitating condition that, when not effectively treated, can result in significant deficits and lifelong dependency. Common issues like neglecting the use of limbs can lead to weakness in chronic stroke cases. In rehabilitation, active patient participation significantly activates the sensorimotor network during motor control, unlike passive movement. This study aims to assess how electromyographic triggering (EMG-triggered) robotic treatments affect walking, ankle muscle force after an ischemic stroke, and the coactivation of agonist and antagonist muscles, which contributes to neuroplasticity with the assistance of biofeedback using robotics. Methods: The study utilized robotic techniques based on electromyography (EMG) for daily rehabilitation in long-term stroke patients, offering feedback and monitoring progress. Each patient received one session per day for two weeks, with the intervention group undergoing 45 minutes of robot-assisted training and exercise at the hospital, while the control group performed exercises at home. Eight participants with impaired motor function and gait after stroke were involved in the study. EMG-based biofeedback exercises were administered through the LUNA neuro-robotic machine, progressing from trigger and release mode to trigger and hold, and later transitioning to dynamic mode. Assessments were conducted at baseline and after two weeks, including the Timed Up and Go (TUG) test, a 10-meter walk test (10m), Berg Balance Scale (BBG), and gait parameters like cadence, step length, upper limb strength measured by EMG threshold in microvolts, and force in Newton meters. Results: The study utilized a scale to assess motor strength and balance, illustrating the benefits of EMG-biofeedback following LUNA robotic therapy. In the analysis of the left hemiparetic group, an increase in strength post-rehabilitation was observed. The pre-TUG mean value was 72.4, which decreased to 42.4 ± 0.03880133 seconds post-rehabilitation, with a significant difference indicated by a p-value below 0.05, reflecting a reduced task completion time. Similarly, in the force-based task, the pre-knee dynamic force in Newton meters was 18.2NM, which increased to 31.26NM during knee extension post-rehabilitation. The post-student t-test showed a p-value of 0.026, signifying a significant difference. This indicated an increase in the strength of knee extensor muscles after LUNA robotic rehabilitation. Lastly, at baseline, the EMG value for ankle dorsiflexion was 5.11 (µV), which increased to 43.4 ± 0.06 µV post-rehabilitation, signifying an increase in the threshold and the patient's ability to generate more motor units during left ankle dorsiflexion. Conclusion: This study aimed to evaluate the impact of EMG and dynamic force-based rehabilitation devices on walking and strength of the affected side in chronic stroke patients without nominal data comparisons among stroke patients. Additionally, it provides insights into the inclusion of EMG-triggered neurorehabilitation robots in the daily rehabilitation of patients.

Keywords: neurorehabilitation, robotic therapy, stroke, strength, paralysis

Procedia PDF Downloads 39

Authors: Ali Kazemi

Abstract:

Those within the volatile and interconnected international economic markets, appropriately predicting market trends, hold substantial fees for traders and financial establishments. Traditional device mastering strategies have made full-size strides in forecasting marketplace movements; however, monetary data's complicated and networked nature calls for extra sophisticated processes. This observation offers a groundbreaking method for monetary marketplace prediction that leverages the synergistic capability of Graph Convolutional Networks (GCNs) and Long Short-Term Memory (LSTM) networks. Our suggested algorithm is meticulously designed to forecast the traits of inventory market indices and cryptocurrency costs, utilizing a comprehensive dataset spanning from January 1, 2015, to December 31, 2023. This era, marked by sizable volatility and transformation in financial markets, affords a solid basis for schooling and checking out our predictive version. Our algorithm integrates diverse facts to construct a dynamic economic graph that correctly reflects market intricacies. We meticulously collect opening, closing, and high and low costs daily for key inventory marketplace indices (e.g., S&P 500, NASDAQ) and widespread cryptocurrencies (e.g., Bitcoin, Ethereum), ensuring a holistic view of marketplace traits. Daily trading volumes are also incorporated to seize marketplace pastime and liquidity, providing critical insights into the market's shopping for and selling dynamics. Furthermore, recognizing the profound influence of the monetary surroundings on financial markets, we integrate critical macroeconomic signs with hobby fees, inflation rates, GDP increase, and unemployment costs into our model. Our GCN algorithm is adept at learning the relational patterns amongst specific financial devices represented as nodes in a comprehensive market graph. Edges in this graph encapsulate the relationships based totally on co-movement styles and sentiment correlations, enabling our version to grasp the complicated community of influences governing marketplace moves. Complementing this, our LSTM algorithm is trained on sequences of the spatial-temporal illustration discovered through the GCN, enriched with historic fee and extent records. This lets the LSTM seize and expect temporal marketplace developments accurately. Inside the complete assessment of our GCN-LSTM algorithm across the inventory marketplace and cryptocurrency datasets, the version confirmed advanced predictive accuracy and profitability compared to conventional and opportunity machine learning to know benchmarks. Specifically, the model performed a Mean Absolute Error (MAE) of 0.85%, indicating high precision in predicting day-by-day charge movements. The RMSE was recorded at 1.2%, underscoring the model's effectiveness in minimizing tremendous prediction mistakes, which is vital in volatile markets. Furthermore, when assessing the model's predictive performance on directional market movements, it achieved an accuracy rate of 78%, significantly outperforming the benchmark models, averaging an accuracy of 65%. This high degree of accuracy is instrumental for techniques that predict the course of price moves. This study showcases the efficacy of mixing graph-based totally and sequential deep learning knowledge in economic marketplace prediction and highlights the fee of a comprehensive, records-pushed evaluation framework. Our findings promise to revolutionize investment techniques and hazard management practices, offering investors and economic analysts a powerful device to navigate the complexities of cutting-edge economic markets.

Keywords: financial market prediction, graph convolutional networks (GCNs), long short-term memory (LSTM), cryptocurrency forecasting

Procedia PDF Downloads 29

Authors: Mariza Kaskara, Stella Girtsou, Maria Prodromou, Alexia Tsouni, Christodoulos Mettas, Stavroula Alatza, Kyriaki Fotiou, Marios Tzouvaras, Charalampos Kontoes, Diofantos Hadjimitsis

Abstract:

Natural hazard assessment and monitoring are crucial components in managing the risks associated with fires, floods, and geohazards, particularly in regions prone to these natural disasters, such as Greece and Cyprus. Recent advancements in technology led to the development of state-of-the-art systems for assessing and monitoring these hazards. These technologies, developed by the BEYOND Center of Excellence of the National Observatory of Athens, have been successfully applied in Greece and are now set to be transferred to Cyprus. The implementation of these advanced technologies in Greece has significantly improved the country's ability to respond to these natural hazards. Enhancing disaster resilience is crucial as it significantly improves our ability to predict, prepare for, and mitigate the impacts of natural disasters, ultimately saving lives and reducing economic losses. For wildfire risk assessment, a scalar wildfire occurrence risk index has been created based on the predictions of machine learning models. Our objective was to train an ML model that learns to derive a fire susceptibility score when given as input a vector of features assigned to certain spatiotemporal coordinates. Predicting fire danger is crucial for the sustainable management of forest fires as it provides essential information for designing effective prevention measures and facilitating response planning for potential fire incidents. For flood risk assessment, a multi-faceted approach has been employed, including the application of remote sensing techniques, the collection and processing of data from population, buildings, technical studies and field visits, as well as hydrological and hydraulic simulations. All input data are used to create precise flood hazard maps according to various flooding scenarios, detailed flood vulnerability and flood exposure maps, which finally produce the flood risk map. Critical points are identified, and mitigation measures are proposed for the worst-case scenario, namely, refuge areas are defined, and escape routes are designed. Flood risk maps can assist in raising awareness and save lives. For geohazards monitoring (e.g., landslides, subsidence), synthetic aperture radar (SAR) and optical satellite imagery have been combined with geomorphological and meteorological data and other landslide/ground deformation contributing factors. To monitor critical infrastructures, including dams, advanced InSAR (Interferometric SAR) methodologies are used for identifying surface movements through time. Monitoring these hazards provides valuable information for understanding processes and could lead to early warning systems to protect people and infrastructure. The success of these systems in Greece has paved the way for their transfer to Cyprus to enhance Cyprus's capabilities in natural hazard assessment and monitoring. This transfer is being made through knowledge transfer activities, fostering continuous collaboration between Greek and Cypriot experts. Furthermore, small demonstration actions are implemented to showcase the effectiveness of these technologies in real-world scenarios. In conclusion, the transfer of advanced natural hazard assessment technologies from Greece to Cyprus represents a significant step forward in enhancing the entire region's resilience to disasters. The EXCELSIOR project, funding this opportunity, is committed to empowering Cyprus with the tools and expertise needed to effectively manage and mitigate the risks associated with these natural hazards. Acknowledgment: Authors acknowledge the 'EXCELSIOR': ERATOSTHENES: Excellence Research Centre for Earth Surveillance and Space-Based Monitoring of the Environment H2020 Widespread Teaming project.

Keywords: earth observation, monitoring, natural hazards, remote sensing

Procedia PDF Downloads 13