Search results for: Nakamura Yoshio

Authors: Takashi Nakamura

Abstract:

In the Tokyo metropolitan area, traffic congestion and long commute times are caused by overconcentration in the central area. To resolve these problems, a core business city development policy was adopted in 1988. The core business cities, which include Yokohama, Chiba, Saitama, Tachikawa, and others, have designated business facilities accumulation districts where assistance measures are applied. Focusing on Yokohama city, this study investigates the trends in the number of offices, employees, and commuters at 2001 and 2012 Economic Census, as well as the average commute time in the Tokyo metropolitan area from 2005 to 2015 Metropolitan Transportation Census. Surveys were administered in 2001 and 2012 Economic Census to participants who worked in Yokohama, according to their distribution in the city's 1,757 subregions. Four main findings emerged: (1) The number of offices increased in Yokohama when the number of offices decreased in the Tokyo metropolitan area overall. Additionally, the number of employees at Yokohama increased. (2) The number of commuters to Tokyo's central area increased from Saitama prefecture, Tokyo Tama area, and Tokyo central area. However, it decreased from other areas. (3) The average commute time in the Tokyo metropolitan area was 67.7 minutes in 2015, a slight decrease from 2005 and 2010. (4) The number of employees at business facilities accumulation districts in Yokohama city increased greatly.

Keywords: core business city development policy, commute time, number of employees, Yokohama city, distribution of employees

Procedia PDF Downloads 114

Authors: Norimichi Nagano, Yoshio Hirano, Tsutomu Yasukawa

Abstract:

Mesenchymal stem cells are thought to confer neuroprotection, facilitate tissue regeneration and exert their effects on retinal degenerative diseases, however, adverse events such as proliferative vitreoretinopathy and preretinal membrane disease associated with cell suspension transplantation have also been reported. We have recently developed human (allogeneic) adipose tissue-derived mesenchymal stem cell (adMSC) sheets through our proprietary sheet transformation technique, which could potentially mitigate these adverse events. To clarify the properties of our adMSC sheets named PAL-222, we performed in vitro studies such as viability testing, cytokine secretions by ELISA, immunohistochemical study, and migration assay. The viability of the cells exceeded 70%. Vascular Endothelial Growth Factor (VEGF) and Pigment Epithelium-Derived Factor (PEDF), which are quite important cytokines for the retinal area, were observed. PAL-222 expressed type I collagen, a strength marker, type IV collagen, a marker of the basement membrane, and elastin, an elasticity marker. Finally, the migration assay was performed and showed negative, which means that PAL-222 is stably kept in the topical area and does not come to pieces. Next, to evaluate the efficacy in vivo, we transplanted PAL-222 into the subretinal space of the eye of Royal College of Surgeons rats with congenital retinal degeneration and assessed it for three weeks after transplantation. We confirmed that PAL-222 suppressed the decrease in the thickness of the outer nuclear layer, which means that the photoreceptor protective effect treated with PAL-222 was significantly higher than that in the sham group. (p < 0.01). This finding demonstrates that PAL-222 showed their retinoprotective effect in a model of congenital retinal degeneration. As the study suggested the efficacy of PAL-222 in both in vitro and in vivo studies, we are presently engaged in clinical trials of PAL-222 for myopic chorioretinal atrophy, which is one of the retinal degenerative diseases, for the purpose of regenerative medicine.

Keywords: cell sheet, clinical trial, mesenchymal stem cell, myopic chorioretinal atrophy

Procedia PDF Downloads 50

Authors: Sayuri Inoue, Naohiro Nakamura, Tsubasa Hamada

Abstract:

The seismic design method of buildings is classified into two types: static design and dynamic design. The static design is a design method that exerts static force as seismic force and is a relatively simple design method created based on the experience of seismic motion in the past 100 years. At present, static design is used for most of the Japanese buildings. Dynamic design mainly refers to the time history response analysis. It is a comparatively difficult design method that input the earthquake motion assumed in the building model and examine the response. Currently, it is only used for skyscrapers and specific buildings. In the present design standard in Japan, it is good to use either the design method of the static design and the dynamic design in the medium and high-rise buildings. However, when actually designing middle and high-rise buildings by two kinds of design methods, the relatively simple static design method satisfies the criteria, but in the case of a little difficult dynamic design method, the criterion isn't often satisfied. This is because the dynamic design method was built with the intention of designing super high-rise buildings. In short, higher safety is required as compared with general buildings, and criteria become stricter. The authors consider applying the dynamic design method to general buildings designed by the static design method so far. The reason is that application of the dynamic design method is reasonable for buildings that are out of the conventional standard structural form such as emphasizing design. For the purpose, it is important to compare the design results when the criteria of both design methods are arranged side by side. In this study, we performed time history response analysis to medium-rise buildings that were actually designed with allowable stress method. Quantitative comparison between static design and dynamic design was conducted, and characteristics of both design methods were examined.

Keywords: buildings, seismic design, allowable stress design, time history response analysis, Japanese seismic code

Procedia PDF Downloads 134

Authors: Fiona S. Motswaiso, Kengo Nakamura, Takeshi Komai

Abstract:

Heavy metals may occur naturally in rocks and soils, but elevated quantities of them are being gradually released into the environment by anthropogenic activities such as mining. In order to address issues of heavy metal water and soil pollution, a distinction needs to be made between natural and anthropogenic anomalies. The current study aims at characterizing the spatial distribution of trace elements and evaluate site-specific geochemical background concentrations of trace elements in the mine soils examined, and also to discriminate between lithogenic and anthropogenic sources of enrichment around a copper-nickel mining town in Selibe-Phikwe, Botswana. A total of 20 Soil samples, 11 river sediment, and 9 river water samples were collected from an area of 625m² within the precincts of the mine and the smelter. The concentrations of metals (Cu, Ni, Pb, Zn, Cr, Ni, Mn, As, Pb, and Co) were determined by using an ICP-MS after digestion with aqua regia. Major elements were also determined using ED-XRF. Water pH and EC were measured on site and recorded while soil pH and EC were also determined in the laboratory after performing water elution tests. The highest Cu and Ni concentrations in soil are 593mg/kg and 453mg/kg respectively, which is 3 times higher than the crustal composition values and 2 times higher than the South African minimum allowable levels of heavy metals in soils. The level of copper contamination was higher than that of nickel and other contaminants. Water pH levels ranged from basic (9) to very acidic (3) in areas closer to the mine/smelter. There is high variation in heavy metal concentration, eg. Cu suggesting that some sites depict regional natural background concentrations while other depict anthropogenic sources.

Keywords: contamination, geochemical baseline, heavy metals, soils

Procedia PDF Downloads 127

Authors: Harumi Katayama, Mina Suzuki, Taeko Muramatsu, Yui Shimogawa, Yoshimi Mizushima, Mitsuo Hiramatsu, Kimitsugu Nakamura, Takeshi Suzue

Abstract:

Purpose: There are very often footbaths in the hot-spring area as culture from old days in Japan. This culture moderately supported mental and physical health among people. In Japanese hospitals, nurses provide footbath for severe patients to mental comfortable. However, there are only a few evidences effect of footbath for mental comfortable. In this presentation, we show the effect of stress relief of the footbath using biomarker among 35 college students in volunteer. Methods: The experiment was designed in two groups of the footbath group and the simple relaxation group randomly. As mental load, Kraepelin test was given to the students beforehand. Ultra-weak chemiluminescence (UCL) in saliva and self-administered liner scale measurable emotional state were measured on four times concurrently; there is before and after the mental load, after the stress relief, and 30 minutes after the stress relief. The scale that measured emotional state was consisted of 7 factors; there is excitement, relaxation, vigorous, fatigue, tension, calm, and sleepiness with 22 items. ANOVA was calculated effect of the footbath for stress relief. Results: The level of UCL (photons/100sec) was significantly increased in response on both groups after mental load. After the two types of stress relief, UCL (photons/100sec) of footbath group was significantly decreased compared to simple relaxation group. Score of sleepiness and relaxation were significantly increased after the stress relief in the footbath group than the simple relaxation group. However, score of excitement, vigorous, tension, and calm were exhibit the same degree of decrease after the stress relief on both group. Conclusion: It was suggested that salivary UCL may be a sensitive biomarker for mild stress relief as nursing care. In the future, we will measure using UCL to evaluate as stress relief for inpatients, outpatients, or general public as the subjects.

Keywords: bio-marker, footbath, Japan, stress relief

Procedia PDF Downloads 295

Authors: Fumiya Sugino, Naohiro Nakamura, Yuji Miyazu

Abstract:

In earthquake resistant design of buildings, evaluation of vibration characteristics is important. In recent years, due to the increment of super high-rise buildings, the evaluation of response is important for not only the first mode but also higher modes. The knowledge of vibration characteristics in buildings is mostly limited to the first mode and the knowledge of higher modes is still insufficient. In this paper, using earthquake observation records of a SRC building by applying frequency filter to ARX model, characteristics of first and second modes were studied. First, we studied the change of the eigen frequency and the damping ratio during the 3.11 earthquake. The eigen frequency gradually decreases from the time of earthquake occurrence, and it is almost stable after about 150 seconds have passed. At this time, the decreasing rates of the 1st and 2nd eigen frequencies are both about 0.7. Although the damping ratio has more large error than the eigen frequency, both the 1st and 2nd damping ratio are 3 to 5%. Also, there is a strong correlation between the 1st and 2nd eigen frequency, and the regression line is y=3.17x. In the damping ratio, the regression line is y=0.90x. Therefore 1st and 2nd damping ratios are approximately the same degree. Next, we study the eigen frequency and damping ratio from 1998 after 3.11 earthquakes, the final year is 2014. In all the considered earthquakes, they are connected in order of occurrence respectively. The eigen frequency slowly declined from immediately after completion, and tend to stabilize after several years. Although it has declined greatly after the 3.11 earthquake. Both the decresing rate of the 1st and 2nd eigen frequencies until about 7 years later are about 0.8. For the damping ratio, both the 1st and 2nd are about 1 to 6%. After the 3.11 earthquake, the 1st increases by about 1% and the 2nd increases by less than 1%. For the eigen frequency, there is a strong correlation between the 1st and 2nd, and the regression line is y=3.17x. For the damping ratio, the regression line is y=1.01x. Therefore, it can be said that the 1st and 2nd damping ratio is approximately the same degree. Based on the above results, changes in eigen frequency and damping ratio are summarized as follows. In the long-term study of the eigen frequency, both the 1st and 2nd gradually declined from immediately after completion, and tended to stabilize after a few years. Further it declined after the 3.11 earthquake. In addition, there is a strong correlation between the 1st and 2nd, and the declining time and the decreasing rate are the same degree. In the long-term study of the damping ratio, both the 1st and 2nd are about 1 to 6%. After the 3.11 earthquake, the 1st increases by about 1%, the 2nd increases by less than 1%. Also, the 1st and 2nd are approximately the same degree.

Keywords: eigenfrequency, damping ratio, ARX model, earthquake observation records

Procedia PDF Downloads 191

Authors: Waheed Ahmad Safi, Shunichi Nakamura

Abstract:

Concrete filled steel I-girder (CFIG) bridge was proposed and the bending and shear strength was confirmed by experiments. The area surrounded by the upper and lower flanges and the web is filled with concrete in CFIG, which is used to the intermediate support of a continuous girder. Three-dimensional finite element models were established to simulate the bending and shear behaviors of CFIG and to clarify the load transfer mechanism. Steel plates and filled concrete were modeled as a three-dimensional 8-node solid element and steel reinforcement bars as a three-dimensional 2-node truss element. The elements were mostly divided into the 50 x 50 mm mesh size. The non-linear stress-strain relation is assumed for concrete in compression including the softening effect after the peak, and the stress increases linearly for concrete in tension until concrete cracking but then decreases due to tension stiffening effect. The stress-strain relation for steel plates was tri-linear and that for reinforcements was bi-linear. The concrete and the steel plates were rigidly connected. The developed FEM model was applied to simulate and analysis the bending behaviors of the CFIG specimens. The vertical displacements and the strains of steel plates and the filled concrete obtained by FEM agreed very well with the test results until the yield load. The specimens collapsed when the upper flange buckled or the concrete spalled off. These phenomena cannot be properly analyzed by FEM, which produces a small discrepancy at the ultimate states. The FEM model was also applied to simulate and analysis the shear tests of the CFIG specimens. The vertical displacements and strains of steel and concrete calculated by FEM model agreed well with the test results. A truss action was confirmed by the FEM and the experiment, clarifying that shear forces were mainly resisted by the tension strut of the steel plate and the compression strut of the filled concrete acting in the diagonal direction. A trail design with the CFIG was carried out for a four-span continuous highway bridge and the design method was established. Construction cost was estimated about 12% lower than that of a conventional steel I-section girder.

Keywords: concrete filled steel I-girder, bending strength, FEM, limit states design, steel I-girder, shear strength

Procedia PDF Downloads 190

Authors: Naohiro Nakamura, Takuya Kinoshita, Hiroshi Fukuyama

Abstract:

The 2011 off The Pacific Coast of Tohoku Earthquake caused considerable damage to wide areas of eastern Japan. A large number of earthquake observation records were obtained at various places. To design more earthquake-resistant buildings and improve earthquake disaster prevention, it is necessary to utilize these data to analyze and evaluate the behavior of a building during an earthquake. This paper presents an earthquake response simulation analysis (hereafter a seismic response analysis) that was conducted using data recorded during the main earthquake (hereafter the main shock) as well as the earthquakes before and after it. The data were obtained at a high-rise steel-reinforced concrete (SRC) building in the bay area of Tokyo. We first give an overview of the building, along with the characteristics of the earthquake motion and the building during the main shock. The data indicate that there was a change in the natural period before and after the earthquake. Next, we present the results of our seismic response analysis. First, the analysis model and conditions are shown, and then, the analysis result is compared with the observational records. Using the analysis result, we then study the effect of soil-structure interaction on the response of the building. By identifying the characteristics of the building during the earthquake (i.e., the 1st natural period and the 1st damping ratio) by the Auto-Regressive eXogenous (ARX) model, we compare the analysis result with the observational records so as to evaluate the accuracy of the response analysis. In this study, a lumped-mass system SR model was used to conduct a seismic response analysis using observational data as input waves. The main results of this study are as follows: 1) The observational records of the 3/11 main shock put it between a level 1 and level 2 earthquake. The result of the ground response analysis showed that the maximum shear strain in the ground was about 0.1% and that the possibility of liquefaction occurring was low. 2) During the 3/11 main shock, the observed wave showed that the eigenperiod of the building became longer; this behavior could be generally reproduced in the response analysis. This prolonged eigenperiod was due to the nonlinearity of the superstructure, and the effect of the nonlinearity of the ground seems to have been small. 3) As for the 4/11 aftershock, a continuous analysis in which the subject seismic wave was input after the 3/11 main shock was input was conducted. The analyzed values generally corresponded well with the observed values. This means that the effect of the nonlinearity of the main shock was retained by the building. It is important to consider this when conducting the response evaluation. 4) The first period and the damping ratio during a vibration were evaluated by an ARX model. Our results show that the response analysis model in this study is generally good at estimating a change in the response of the building during a vibration.

Keywords: ARX model, response analysis, SRC building, the 2011 off the Pacific Coast of Tohoku Earthquake

Procedia PDF Downloads 138

Authors: Kuniko Muramoto, Tadashi Nakamura, Shiro Dan

Abstract:

On March 11, 2011 the Great East Japan Disaster caused widespread damage. In the aftermath, we searched for ways to provide long-term support and enhanced resilience to affected areas, arriving at the Family Manga Exhibition: an art collection portraying family life. It became a tool for community outreach and intervention, and we implemented support programs by collaborating with local support agencies. This 10-year project has been touring through four prefectures in Tohoku since the disaster struck, bearing witness to the effects of disaster and recovery alike. At this five-year mark, we use a narrative perspective to present our findings and reflect on post-disaster community support. It is important to note that the exhibition’s art does not directly depict the disaster; it portrays stories of anonymous families instead. They stimulate viewers’ memories and remind them of their own family stories. We analyzed viewers' oral and written responses to the exhibition and discovered that family manga as an art form enhances the viewer’s sense of connection to people close to them. We also discovered that the viewers gained more universal perspective on their own situations by viewing the exhibition. Manga, we found, offered a certain safety by enabling the viewers to control how they would interact with the exhibition's content and themes. In addition, the purpose of the project was for us to become witnesses of the disaster and recovery. Supporters of the project became active listeners, functioning as interactive agents who helped forming stories. Voices of the story tellers and the listeners layered upon each other and, as a result, converged into brand new narratives. The essence of traumatic experience is ‘the sense of overwhelming powerlessness and isolation’. When we redefine trauma as ‘broken relationships’, we can say that ‘enhancing relationships’ and ‘weaving relationships’ are what strengthen our resilience. This project used narrative as a modality to fortify the resilience of people involved by enhancing the social capital of bonding, bridging, and linking. The manga exhibition functioned as a tool to achieve this end, suggesting that similar applications are possible. Programs we held in-between manga exhibitions also served to enhance narratives of resiliency in the regions. However, we will save that story for another time. We hope to continue collecting the precious and polyphonic voices of people to present as stories born out of the Great East Japan Disaster. This effort extends beyond the immediately affected area by helping us prepare our resilience for future disasters.

Keywords: community, manga, narrative, resilience

Procedia PDF Downloads 185

Authors: Shinji Ukita, Naohiro Nakamura, Yuji Miyazu

Abstract:

In considering the dynamic characteristics of structure, natural frequency and damping ratio are useful indicator. When performing dynamic design, it's necessary to select an appropriate initial damping model and hysteretic model. In the linear region, the setting of initial damping model influences the response, and in the nonlinear region, the combination of initial damping model and hysteretic model influences the response. However, the dynamic characteristics of structure in the nonlinear region remain unclear. In this paper, we studied the effect of setting of initial damping model and hysteretic model on the dynamic characteristics of structure. On initial damping model setting, Initial stiffness proportional, Tangent stiffness proportional, and Rayleigh-type were used. On hysteretic model setting, TAKEDA model and Normal-trilinear model were used. As a study method, dynamic analysis was performed using a lumped mass model of base-fixed. During analysis, the maximum acceleration of input earthquake motion was gradually increased from 1 to 600 gal. The dynamic characteristics were calculated using the ARX model. Then, the characteristics of 1st and 2nd natural frequency and 1st damping ratio were evaluated. Input earthquake motion was simulated wave that the Building Center of Japan has published. On the building model, an RC building with 30×30m planes on each floor was assumed. The story height was 3m and the maximum height was 18m. Unit weight for each floor was 1.0t/m2. The building natural period was set to 0.36sec, and the initial stiffness of each floor was calculated by assuming the 1st mode to be an inverted triangle. First, we investigated the difference of the dynamic characteristics depending on the difference of initial damping model setting. With the increase in the maximum acceleration of the input earthquake motions, the 1st and 2nd natural frequency decreased, and the 1st damping ratio increased. Then, in the natural frequency, the difference due to initial damping model setting was small, but in the damping ratio, a significant difference was observed (Initial stiffness proportional≒Rayleigh type>Tangent stiffness proportional). The acceleration and the displacement of the earthquake response were largest in the tangent stiffness proportional. In the range where the acceleration response increased, the damping ratio was constant. In the range where the acceleration response was constant, the damping ratio increased. Next, we investigated the difference of the dynamic characteristics depending on the difference of hysteretic model setting. With the increase in the maximum acceleration of the input earthquake motions, the natural frequency decreased in TAKEDA model, but in Normal-trilinear model, the natural frequency didn’t change. The damping ratio in TAKEDA model was higher than that in Normal-trilinear model, although, both in TAKEDA model and Normal-trilinear model, the damping ratio increased. In conclusion, in initial damping model setting, the tangent stiffness proportional was evaluated the most. In the hysteretic model setting, TAKEDA model was more appreciated than the Normal-trilinear model in the nonlinear region. Our results would provide useful indicator on dynamic design.

Keywords: initial damping model, damping ratio, dynamic analysis, hysteretic model, natural frequency

Procedia PDF Downloads 153

Authors: Siddhant Sethi, Yasuharu Takashima, Shigetaka Nakamura, Kenzo Fujimoto

Abstract:

Site-directed mutagenesis is a renowned technique to introduce specific mutations in the genome. To achieve site-directed mutagenesis, many chemical and enzymatic approaches have been reported in the past like disulphite induced genome editing, CRISPR-Cas9, TALEN etc. The chemical methods are invasive whereas the enzymatic approaches are time-consuming and expensive. Most of these techniques are unusable in the cellular application due to their toxicity and other limitations. Photo-chemical cytosine deamination, introduced in 2010, is one of the major technique for enzyme-free single-point mutation of cytosine to uracil in DNA and RNA, wherein, 3-cyanovinylcarbazole nucleoside (CNVK) containing oligodeoxyribonucleotide (ODN) having CNVK at -1 position to that of target cytosine is reversibly crosslinked to target DNA strand using 366 nm and then incubated at 90ºC to accommodate deamination. This technique is superior to enzymatic methods of site-directed mutagenesis but has a disadvantage that it requires the use of high temperature for the deamination step which restricts its applicability in the in vivo applications. This study has been focused on improving the technique by reducing the temperature required for deamination. Firstly, the photo-cross-linker, CNVK has been modified by replacing cyano group attached to vinyl group with methyl ester (OMeVK), amide (NH2VK), and carboxylic acid (OHVK) to observe the acceleration in the deamination of target cytosine cross-linked to vinylcarbazole derivative. Among the derivatives, OHVK has shown 2 times acceleration in deamination reaction as compared to CNVK, while the other two derivatives have shown deceleration towards deamination reaction. The trend of rate of deamination reaction follows the same order as that of hydrophilicity of the vinylcarbazole derivatives. OHVK being most hydrophilic has shown highest acceleration while OMeVK is least hydrophilic has proven to be least active for deamination. Secondly, in the related study, the counter-base of the target cytosine, guanine has been replaced by inosine, 2-aminopurine, nebularine, and 5-nitroindole having distinct hydrogen bonding patterns with target cytosine. Among the ODNs with these counter bases, ODN with inosine has shown 12 fold acceleration towards deamination of cytosine cross-linked to CNVK at physiological conditions as compared to guanosine. Whereas, when 2-aminopurine, nebularine, and 5-nitroindole were used, no deamination reaction took place. It can be concluded that inosine has potential to be used as the counter base of target cytosine for the CNVK mediated photo-cross-linking induced deamination of cytosine. The increase in rate of deamination reaction has been attributed to pattern and number of hydrogen bonding between the cytosine and counter base. One of the important factor is presence of hydrogen bond between exo-cyclic amino group of cytosine and the counter base. These results will be useful for development of more efficient technique for site-directed mutagenesis for C → U transformations in the DNA/RNA which might be used in the living system for treatment of various genetic disorders and genome engineering for making designer and non-native proteins.

Keywords: C to U transformation, DNA editing, genome engineering, ultra-fast photo-cross-linking

Procedia PDF Downloads 207

Authors: Ja-Keoung Koo, Kensuke Nakamura, Hyohun Kim, Dongwha Shin, Yeonseok Kim, Ji-Su Ahn, Byung-Woo Hong

Abstract:

The machine learning techniques based on a convolutional neural network (CNN) have been actively developed and successfully applied to a variety of image analysis tasks including reconstruction, noise reduction, resolution enhancement, segmentation, motion estimation, object recognition. The classical visual information processing that ranges from low level tasks to high level ones has been widely developed in the deep learning framework. It is generally considered as a challenging problem to derive visual interpretation from high dimensional imagery data. A CNN is a class of feed-forward artificial neural network that usually consists of deep layers the connections of which are established by a series of non-linear operations. The CNN architecture is known to be shift invariant due to its shared weights and translation invariance characteristics. However, it is often computationally intractable to optimize the network in particular with a large number of convolution layers due to a large number of unknowns to be optimized with respect to the training set that is generally required to be large enough to effectively generalize the model under consideration. It is also necessary to limit the size of convolution kernels due to the computational expense despite of the recent development of effective parallel processing machinery, which leads to the use of the constantly small size of the convolution kernels throughout the deep CNN architecture. However, it is often desired to consider different scales in the analysis of visual features at different layers in the network. Thus, we propose a CNN model where different sizes of the convolution kernels are applied at each layer based on the random projection. We apply random filters with varying sizes and associate the filter responses with scalar weights that correspond to the standard deviation of the random filters. We are allowed to use large number of random filters with the cost of one scalar unknown for each filter. The computational cost in the back-propagation procedure does not increase with the larger size of the filters even though the additional computational cost is required in the computation of convolution in the feed-forward procedure. The use of random kernels with varying sizes allows to effectively analyze image features at multiple scales leading to a better generalization. The robustness and effectiveness of the proposed CNN based on random kernels are demonstrated by numerical experiments where the quantitative comparison of the well-known CNN architectures and our models that simply replace the convolution kernels with the random filters is performed. The experimental results indicate that our model achieves better performance with less number of unknown weights. The proposed algorithm has a high potential in the application of a variety of visual tasks based on the CNN framework. Acknowledgement—This work was supported by the MISP (Ministry of Science and ICT), Korea, under the National Program for Excellence in SW (20170001000011001) supervised by IITP, and NRF-2014R1A2A1A11051941, NRF2017R1A2B4006023.

Keywords: deep learning, convolutional neural network, random kernel, random projection, dimensionality reduction, object recognition

Procedia PDF Downloads 257

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 165