Search results for: Mala William Armand

Authors: William Michael Short

Abstract:

Semantic Web’ technologies such as the Unified Medical Language System Metathesaurus, SNOMED-CT, and MeSH have been touted as transformational for the way users access online medical and health information, enabling both the automated analysis of natural-language data and the integration of heterogeneous healthrelated resources distributed across the Internet through the use of standardized terminologies that capture concepts and relationships between concepts that are expressed differently across datasets. However, the approaches that have so far characterized ‘semantic bioinformatics’ have not yet fulfilled the promise of the Semantic Web for medical and health information retrieval applications. This paper argues within the perspective of cognitive linguistics and cognitive anthropology that four features of human meaning-making must be taken into account before the potential of semantic technologies can be realized for this domain. First, many semantic technologies operate exclusively at the level of the word. However, texts convey meanings in ways beyond lexical semantics. For example, transitivity patterns (distributions of active or passive voice) and modality patterns (configurations of modal constituents like may, might, could, would, should) convey experiential and epistemic meanings that are not captured by single words. Language users also naturally associate stretches of text with discrete meanings, so that whole sentences can be ascribed senses similar to the senses of words (so-called ‘discourse topics’). Second, natural language processing systems tend to operate according to the principle of ‘one token, one tag’. For instance, occurrences of the word sound must be disambiguated for part of speech: in context, is sound a noun or a verb or an adjective? In syntactic analysis, deterministic annotation methods may be acceptable. But because natural language utterances are typically characterized by polyvalency and ambiguities of all kinds (including intentional ambiguities), such methods leave the meanings of texts highly impoverished. Third, ontologies tend to be disconnected from everyday language use and so struggle in cases where single concepts are captured through complex lexicalizations that involve profile shifts or other embodied representations. More problematically, concept graphs tend to capture ‘expert’ technical models rather than ‘folk’ models of knowledge and so may not match users’ common-sense intuitions about the organization of concepts in prototypical structures rather than Aristotelian categories. Fourth, and finally, most ontologies do not recognize the pervasively figurative character of human language. However, since the time of Galen the widespread use of metaphor in the linguistic usage of both medical professionals and lay persons has been recognized. In particular, metaphor is a well-documented linguistic tool for communicating experiences of pain. Because semantic medical knowledge-bases are designed to help capture variations within technical vocabularies – rather than the kinds of conventionalized figurative semantics that practitioners as well as patients actually utilize in clinical description and diagnosis – they fail to capture this dimension of linguistic usage. The failure of semantic technologies in these respects degrades the efficiency and efficacy not only of medical research, where information retrieval inefficiencies can lead to direct financial costs to organizations, but also of care provision, especially in contexts of patients’ self-management of complex medical conditions.

Keywords: ambiguity, bioinformatics, language, meaning, metaphor, ontology, semantic web, semantics

Procedia PDF Downloads 99

Authors: Mian Huang, Chi Ma, Junyu Lin, William Lu

Abstract:

Introduction: Bone health attracts more attention recently and an intelligent question and answer (QA) chatbot for osteoporosis is helpful for science popularization. With Generative Pretrained Transformer (GPT) technology developing, we build an osteoporosis corpus dataset and then fine-tune LLaMA, a famous open-source GPT foundation large language model(LLM), on our self-constructed osteoporosis corpus. Evaluated by clinical orthopedic experts, our fine-tuned model outperforms vanilla LLaMA on osteoporosis QA task in Chinese. Three-dimensional quantitative computed tomography (QCT) measured bone mineral density (BMD) is considered as more accurate than DXA for BMD measurement in recent years. We develop an automatic Phantom-less QCT(PL-QCT) that is more efficient for BMD measurement since no need of an external phantom for calibration. Combined with LLM on osteoporosis, our PL-QCT provides efficient and accurate BMD measurement for our chatbot users. Material and Methods: We build an osteoporosis corpus containing about 30,000 Chinese literatures whose titles are related to osteoporosis. The whole process is done automatically, including crawling literatures in .pdf format, localizing text/figure/table region by layout segmentation algorithm and recognizing text by OCR algorithm. We train our model by continuous pre-training with Low-rank Adaptation (LoRA, rank=10) technology to adapt LLaMA-7B model to osteoporosis domain, whose basic principle is to mask the next word in the text and make the model predict that word. The loss function is defined as cross-entropy between the predicted and ground-truth word. Experiment is implemented on single NVIDIA A800 GPU for 15 days. Our automatic PL-QCT BMD measurement adopt AI-associated region-of-interest (ROI) generation algorithm for localizing vertebrae-parallel cylinder in cancellous bone. Due to no phantom for BMD calibration, we calculate ROI BMD by CT-BMD of personal muscle and fat. Results & Discussion: Clinical orthopaedic experts are invited to design 5 osteoporosis questions in Chinese, evaluating performance of vanilla LLaMA and our fine-tuned model. Our model outperforms LLaMA on over 80% of these questions, understanding ‘Expert Consensus on Osteoporosis’, ‘QCT for osteoporosis diagnosis’ and ‘Effect of age on osteoporosis’. Detailed results are shown in appendix. Future work may be done by training a larger LLM on the whole orthopaedics with more high-quality domain data, or a multi-modal GPT combining and understanding X-ray and medical text for orthopaedic computer-aided-diagnosis. However, GPT model gives unexpected outputs sometimes, such as repetitive text or seemingly normal but wrong answer (called ‘hallucination’). Even though GPT give correct answers, it cannot be considered as valid clinical diagnoses instead of clinical doctors. The PL-QCT BMD system provided by Bone’s QCT(Bone’s Technology(Shenzhen) Limited) achieves 0.1448mg/cm2(spine) and 0.0002 mg/cm2(hip) mean absolute error(MAE) and linear correlation coefficient R2=0.9970(spine) and R2=0.9991(hip)(compared to QCT-Pro(Mindways)) on 155 patients in three-center clinical trial in Guangzhou, China. Conclusion: This study builds a Chinese osteoporosis corpus and develops a fine-tuned and domain-adapted LLM as well as a PL-QCT BMD measurement system. Our fine-tuned GPT model shows better capability than LLaMA model on most testing questions on osteoporosis. Combined with our PL-QCT BMD system, we are looking forward to providing science popularization and early morning screening for potential osteoporotic patients.

Keywords: GPT, phantom-less QCT, large language model, osteoporosis

Procedia PDF Downloads 36