检索结果-内蒙古大学图书馆

arXiv 2023年

作者： Hu, Yan Chen, Qingyu Du, Jingcheng Peng, Xueqing Keloth, Vipina Kuttichi Zuo, Xu Zhou, Yujia Li, Zehan Jiang, Xiaoqian Lu, Zhiyong Roberts, Kirk Xu, Hua School of Biomedical Informatics University of Texas Health Science at Houston Houston United States Section of Biomedical Informatics and Data Science School of Medicine Yale University New Haven United States National Center for Biotechnology Information National Library of Medicine National Institutes of Health MD United States

Objective: This study quantifies the capabilities of GPT-3.5 and GPT-4 for clinical named entity recognition (NER) tasks and proposes task-specific prompts to improve their performance. Materials and Methods: We evaluated these models on two clinical NER tasks: (1) to extract medical problems, treatments, and tests from clinical notes in the MTSamples corpus, following the 2010 i2b2 concept extraction shared task, and (2) identifying nervous system disorder-related adverse events from safety reports in the vaccine adverse event reporting system (VAERS). To improve the GPT models' performance, we developed a clinical task-specific prompt framework that includes (1) baseline prompts with task description and format specification, (2) annotation guideline-based prompts, (3) error analysis-based instructions, and (4) annotated samples for few-shot learning. We assessed each prompt's effectiveness and compared the models to BioClinicalBERT. Results: Using baseline prompts, GPT-3.5 and GPT-4 achieved relaxed F1 scores of 0.634, 0.804 for MTSamples, and 0.301, 0.593 for VAERS. Additional prompt components consistently improved model performance. When all four components were used, GPT-3.5 and GPT-4 achieved relaxed F1 socres of 0.794, 0.861 for MTSamples and 0.676, 0.736 for VAERS, demonstrating the effectiveness of our prompt framework. Although these results trail BioClinicalBERT (F1 of 0.901 for the MTSamples dataset and 0.802 for the VAERS), it is very promising considering few training samples are needed. Conclusion: While direct application of GPT models to clinical NER tasks falls short of optimal performance, our task-specific prompt framework, incorporating medical knowledge and training samples, significantly enhances GPT models' feasibility for potential clinical applications. Copyright © 2023, The Authors. All rights reserved.

关键词： Medical problems

来源：评论

学校读者我要写书评

暂无评论

A Variational Autoencoder-Feed Forward Networks (VAE-FNN) Model for Classifying DNA Barcodes

A Variational Autoencoder-Feed Forward Networks (VAE-FNN) Mo...

引用

Advances in Electrical, Electronics and Computational Intelligence (ICAEECI), International Conference on

作者： Abhivyakti Yadav Divya S I R Oviya Kalpana Raja Department of Computer Science and Engineering (Artificial Intelligence) Amrita School of Computing Amrita Vishwa Vidyapetham Chennai India Section for Biomedical Informatics and Data Science School of Medicine Yale University New Haven Connecticut USA

DNA Barcodes, which are particular fragments derived from brief sections of DNA (such as mitochondrial, nuclear, and plastid sequences), can be used to identify organisms from the major life kingdoms. In addition to supporting conventional taxonomic techniques, DNA barcoding is a potent tool that advances our knowledge of species diversity and their ecological functions. On a variety of organisms, the use of this approach for species categorization has been successful. In this paper, we examine how DNA barcoding has been used to classify species based on DNA barcodes as well as other related research that has been done over the years on the subject. After experimenting with a number of deep learning models, we have propose a Variational Auto Encoder + Feed Forward Neural Network workflow for classifiying species using DNA barcodes. The models have been assessed on the basis of performance factors including accuracy, recall, and precision. COI, rbcL, matK, and ITS are the specific gene sections that have been identified as barcodes. For both simulated and real datasets, the model can attain an average accuracy of greater than 95 percent. This DNA barcoding approach has the ability to simplify DNA barcode-based species identification and serve as a tool for species categorization.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Suicide Phenotyping from Clinical Notes in Safety-Net Psychiatric Hospital Using Multi-Label Classification with Pre-Trained Language Models

arXiv

引用

arXiv 2024年

作者： Li, Zehan Hu, Yan Lane, Scott Selek, Salih Shahani, Lokesh Machado-Vieira, Rodrigo Soares, Jair Xu, Hua Liu, Hongfang Huang, Ming MacWilliam School of Biomedical Informatics The University of Texas Health Science Center HoustonTX United States Department of Psychiatry & Behavioral Sciences McGovern Medical School The University of Texas Health Science at Houston HoustonTX United States Section of Biomedical Informatics and Data Science School of Medicine Yale University New HavenCT United States

Accurate identification and categorization of suicidal events can yield better suicide precautions, reducing operational burden, and improving care quality in high-acuity psychiatric settings. Pre-trained language models offer promise for identifying suicidality from unstructured clinical narratives. We evaluated the performance of four BERT-based models using two fine-tuning strategies (multiple single-label and single multi-label) for detecting coexisting suicidal events from 500 annotated psychiatric evaluation notes. The notes were labeled for suicidal ideation (SI), suicide attempts (SA), exposure to suicide (ES), and non-suicidal self-injury (NSSI). RoBERTa outperformed other models using binary relevance (acc=0.86, F1=0.78). MentalBERT (F1=0.74) also exceeded BioClinicalBERT (F1=0.72). RoBERTa fine-tuned with a single multi-label classifier further improved performance (acc=0.88, F1=0.81), highlighting that models pre-trained on domain-relevant data and the single multi-label classification strategy enhance efficiency and performance. © 2024, CC BY.

关键词： Classification (of information)

来源：评论

学校读者我要写书评

暂无评论

From reflection to action: Combining machine learning with expert knowledge for nutrition goal recommendations 21

From reflection to action: Combining machine learning with e...

引用

2021 CHI Conference on Human Factors in Computing Systems: Making Waves, Combining Strengths, CHI 2021

作者： Mitchell, Elliot G. Heitkemper, Elizabeth M. Burgermaster, Marissa Levine, Matthew E. Miao, Yishen Hwang, Maria L. Desai, Pooja M. Cassells, Andrea Tobin, Jonathan N. Tabak, Esteban G. Albers, David J. Smaldone, Arlene M. Mamykina, Lena Department of Biomedical Informatics Columbia University United States School of Nursing The University of Texas at Austin United States Department of Population Health Dell Medical School and Department of Nutritional Sciences The University of Texas at Austin United States Department of Computing and Mathematical Sciences California Institute of Technology United States Department of Molecular Cellular and Developmental Biology University of California Santa Barbara United States Department of Science and Math Fashion Institute of Technology United States United States and The Rockefeller University United States Courant Institute of Mathematical Sciences United States University of Colorado Anschutz Medical Campus Section of Informatics and Data Science United States Departments of Pediatrics Biomedical Engineering and Biostatistics and Informatics United States School of Nursing Columbia University United States

ISBN: (纸本)9781450380966

Self-tracking can help personalize self-management interventionsfor chronic conditions like type 2 diabetes (T2D), but refecting onpersonal data requires motivation and literacy. Machine learning(ML) methods can identify patterns, but a key challenge is making actionable suggestions based on personal health *** introduce GlucoGoalie, which combines ML with an expert system to trans-late ML output into personalized nutrition goal suggestions forindividuals with T2D. In a controlled experiment, participants with T2D found that goal suggestions were understandable and action-able. A 4-week in-the-wild deployment study showed that receiving goal suggestions augmented participants' self-discovery, choosing goals highlighted the multifaceted nature of personal preferences, and the experience of following goals demonstrated the importance of feedback and context. However, we identifed tensions between abstract goals and concrete eating experiences and found static text too ambiguous for complex concepts. We discuss implications for ML-based interventions and the need for systems that ofer more interactivity, feedback, and negotiation. © 2021 ACM.

关键词： Feedback

来源：评论

学校读者我要写书评

暂无评论

Improving Entity Recognition Using Ensembles of Deep Learning and Fine-tuned Large Language Models: A Case Study on Adverse Event Extraction from Multiple Sources

arXiv

引用

arXiv 2024年

作者： Li, Yiming Viswaroopan, Deepthi He, William Li, Jianfu Zuo, Xu Xu, Hua Tao, Cui McWilliams School of Biomedical Informatics The University of Texas Health Science Center at Houston HoustonTX77030 United States Department of Electrical & Computer Engineering Pratt School of Engineering Duke University 305 Tower Engineering Building DurhamNC27708 United States Department of Artificial Intelligence and Informatics Mayo Clinic JacksonvilleFL32224 United States Section of Biomedical Informatics and Data Science School of Medicine Yale University New HavenCT06510 United States

Objective Adverse event (AE) extraction following COVID-19 vaccines from text data is crucial for monitoring and analyzing the safety profiles of immunizations, identifying potential risks and ensuring the safe use of these products. Traditional deep learning models are adept at learning intricate feature representations and dependencies in sequential data, but often require extensive labeled data. In contrast, large language models (LLMs) excel in understanding contextual information, but exhibit unstable performance on named entity recognition (NER) tasks, possibly due to their broad but unspecific training. This study aims to evaluate the effectiveness of LLMs and traditional deep learning models in AE extraction, and to assess the impact of ensembling these models on performance. Methods In this study, we utilized reports and posts from the Vaccine Adverse Event Reporting System (VAERS) (n=621), Twitter (n=9,133), and Reddit (n=131) as our corpora. Our goal was to extract three types of entities: vaccine, shot, and adverse event (ae). We explored and fine-tuned (except GPT-4) multiple LLMs, including GPT-2, GPT-3.5, GPT-4, Llama-2 7b, and Llama-2 13b, as well as traditional deep learning models like Recurrent neural network (RNN) and Bidirectional Encoder Representations from Transformers for biomedical Text Mining (BioBERT). To enhance performance, we created ensembles of the three models with the best performance. For evaluation, we used strict and relaxed F1 scores to evaluate the performance for each entity type, and micro-average F1 was used to assess the overall performance. Results The ensemble model achieved the highest performance in "vaccine," "shot," and "ae," with strict F1-scores of 0.878, 0.930, and 0.925, respectively, along with a micro-average score of 0.903. These results underscore the significance of fine-tuning models for specific tasks and demonstrate the effectiveness of ensemble methods in enhancing performance. Conclusion In conclusion,

关键词： Vaccines

来源：评论

学校读者我要写书评

暂无评论

Repurposing Drugs for Alzheimer's Diseases through Link Prediction on biomedical Literature 11

Repurposing Drugs for Alzheimer's Diseases through Link Pred...

引用

11th IEEE International Conference on Healthcare informatics, ICHI 2023

作者： Xiao, Yongkang Hou, Yu Zhou, Huixue Diallo, Gayo Fiszman, Marcelo Wolfson, Julian Kilicoglu, Halil Chen, You Xu, Hua Mantyh, William G. Zhang, Rui University of Minnesota Institute for Health Informatics MinneapolisMN United States University of Minnesota Department of Surgery MinneapolisMN United States University of Bordeaux Inria Sistm Team AHeaD Inserm 1219 Bordeaux Population Health F-33000 France Pontifical Catholic University of Rio de Janeiro Nites - Núcleo de Inovação e Tecnologia em Saúde Brazil University of Minnesota Division of Biostatistics School of Public Health MinneapolisMN United States University of Illinois Urbana-Champaign School of Information Sciences ChampaignIL United States Vanderbilt University Medical Center Department of Biomedical Informatics NashvilleTN United States Yale University Section of Biomedical Informatics and Data Science New HavenCT United States University of Minnesota Department of Neurology MinneapolisMN United States

ISBN: (纸本)9798350302639

Recently, computational drug repurposing has emerged as a promising method for identifying new interventions for diseases. This study predicts novel drugs for Alzheimer's disease (AD) through link prediction on our developed biomedical knowledge graph. We constructed a comprehensive knowledge graph containing AD concepts and various potential interventions, called ADInt, by integrating a dietary supplement (DS) domain knowledge graph, SuppKG, with semantic triples from SemMedDB database. Four knowledge graph embedding models (TransE, RotatE, DistMult and ComplEX) and two graph convolutional network models (R-GCN and CompGCN) were compared to learn the representation of ADInt. R-GCN outperformed other models by evaluating on the time slice test set and the clinical trial test set, and was used to generate the score tables for the link prediction task. According to the results of link prediction, we proposed candidate drugs for AD. In conclusion, we presented a novel methodology to extend an existing knowledge graph and discover novel drugs for AD. Our method can potentially be applied to other clinical problems, such as discovering drug adverse reactions and drug-drug interactions. © 2023 IEEE.

关键词： Knowledge graph

来源：评论

学校读者我要写书评

暂无评论

Delay-induced uncertainty for a paradigmatic glucose-insulin model

arXiv

引用

arXiv 2020年

作者： Karamched, Bhargav Hripcsak, George Albers, David Ott, William Department of Mathematics Institute of Molecular Biophysics Florida State University TallahasseeFL32306 United States Department of Biomedical Informatics Columbia University New YorkNY10032 United States Section of Informatics and Data Science Department of Pediatrics Department of Biomedical Engineering Department of Biostatistics and Informatics University of Colorado Anschutz Medical Campus AuroraCO80045 United States Department of Mathematics University of Houston HoustonTX77204 United States

Medical practice in the intensive care unit is based on the assumption that physiological systems such as the human glucose-insulin system are predictable. We demonstrate that delay within the glucose-insulin system can induce sustained temporal chaos, rendering the system unpredictable. Specifically, we exhibit such chaos for the Ultradian glucose-insulin model. This well-validated, finite-dimensional model represents feedback delay as a three-stage filter. Using the theory of rank one maps from smooth dynamical systems, we precisely explain the nature of the resulting delay-induced uncertainty (DIU). We develop a framework one may use to diagnose DIU in a general oscillatory dynamical system. For infinite-dimensional delay systems, no analog of the theory of rank one maps exists. Nevertheless, we show that the geometric principles encoded in our DIU framework apply to such systems by exhibiting sustained temporal chaos for a linear shear flow. Our results are potentially broadly applicable because delay is ubiquitous throughout mathematical *** Codes 92C50, 92C30, 37N25, 37D25, 37D45, 37G35 Copyright © 2020, The Authors. All rights reserved.

关键词： Insulin

来源：评论

学校读者我要写书评

暂无评论

Kamino: A Scalable Architecture to Support Medical AI Research Using Large Real World data

Kamino: A Scalable Architecture to Support Medical AI Resear...

引用

IEEE International Conference on Healthcare informatics (ICHI)

作者： Fongci Lin Patrick Young Huan He Jimin Huang Roger Gagne Daniel Rice Nathan Price Will Byron Yan Hu Donn Felker Will Button Deniella Meeker Allen Hsiao Hua Xu Charles Torre Wade Schulz Section of Biomedical Informatics and Data Science School of Medicine Yale University New Haven CT USA Department of Laboratory Medicine School of Medicine Yale University New Haven CT USA Digital & Technology Solutions Yale New Haven Hospital New Haven CT USA Mcwilliam School of Biomedical Informatics University of Texas Health Science at Houston Houston TX USA Pediatrics and of Emergency Medicine at Yale School of Medicine Yale University New Haven CT USA

ISBN: (数字)9798350383737

ISBN: (纸本)9798350383744

Electronic Health Records (EHRs) represent a crucial data source for real-world evidence generation. To facilitate biomedical studies using EHRs, standard data models like the OMOP CDM have been developed. Nevertheless, recent advancements in biomedical AI research that leverage EHRs have introduced new challenges, encompassing security considerations, large-scale data retrieval, and computational resource management, including GPUs. This paper introduces Kamino, an innovative architectural solution tailored to support biomedical AI research using EHR data. Kamino offers a user-friendly interface with features designed for efficient team access management in accordance with regulatory requirements. It facilitates direct data retrieval from an OMOP CDM instance and includes a resource allocation system based on Kubernetes orchestration. Here, we demonstrate the practical application and utility of Kamino through a clinical natural language processing task. We firmly believe that such a tool will significantly expedite AI research conducted with EHR data within academic institutions.

关键词： Soft sensors Medical services data models Natural language processing Resource management Security Artificial intelligence

来源：评论

学校读者我要写书评

暂无评论

Use of Intravascular Ultrasound During Peripheral Vascular Interventions for Claudication Is Not Associated with Improved Outcomes

引用

Annals of vascular surgery 2025年

作者： Chen Dun M Libby Weaver Sanuja Bose Katherine M McDermott Sarah E Deery James H Black Jeffrey Siracuse Jesse Columbo Caitlin W Hicks Department of Biomedical Informatics and Data Science The Johns Hopkins University School of Medicine Baltimore MD Department of Surgery The Johns Hopkins University School of Medicine Baltimore MD. Electronic address: cdun1@jhmi.edu. Division of Vascular and Endovascular Surgery University of Virginia Charlottesville VA. Division of Vascular Surgery and Endovascular Therapy The Johns Hopkins University School of Medicine Baltimore MD. Maine Medical Center Portland ME. Division of Acute Care Surgery Department of Surgery Beth Israel Deaconess Medical Center and Harvard Medical School Boston MA. Section of Vascular Surgery Heart and Vascular Center Dartmouth Hitchcock Medical Center Lebanon NH. Division of Vascular and Endovascular Surgery University of Virginia Charlottesville VA. Electronic address: chicks11@jhmi.edu.

INTRODUCTION AND OBJECTIVES:The efficacy of intravascular ultrasound (IVUS) for improving outcomes of peripheral vascular interventions (PVI) has not been well studied. We aimed to evaluate the association of IVUS with long-term outcomes in patients undergoing PVI for claudication. METHODS:We conducted a two-cohort study using data from 100% of Medicare fee-for-service claims (2018-2022) and the Vascular Quality Initiative Vascular Implant Surveillance and Interventional Outcomes Network (VISION; 2016-2019). For both cohorts, we identified all patients who underwent an index (first-time) femoropopliteal PVI for claudication, excluding patients with chronic limb-threatening ischemia (CLTI) and acute limb ischemia. We compared IVUS use over time and by procedure type and setting. We used multivariable Cox proportional hazards models to assess the associations of IVUS with repeat PVI, conversion to CLTI, and amputation. For the Medicare cohort, adjustments were made for baseline patient characteristics, while for the VISION cohort, additional adjustments were made for detailed anatomic factors. All models were clustered by physician. RESULTS:In the Medicare dataset, 69,092 patients (median age 74 years; 40.5% female; 12.1% non-Hispanic Black) underwent an index femoropopliteal PVI for claudication, of whom 22.1% (N=15,253) received IVUS. In the VISION dataset, 6,722 patients (median age 72 years; 38.7% female; 11.6% non-Hispanic Black) underwent an index femoropopliteal PVI for claudication, of whom 3.8% (N=254) received IVUS. The mean follow-up time for both cohorts was 2.7 years. For both the Medicare and VISION cohorts, IVUS use significantly increased over time, particularly in ASC/OBL settings and in conjunction with atherectomy procedures (P<0.001). In the Medicare cohort, IVUS use was associated with a higher hazard of repeat PVI (aHR 1.07, 95% CI 1.02-1.12) and progression to CLTI (aHR 1.11, 95% CI 1.03-1.20) after adjustment compared to PVI without IVUS. In th

关键词： amputation chronic limb-threatening ischemia endovascular reintervention ultrasound

来源：评论

学校读者我要写书评

暂无评论

Relation Extraction Using Large Language Models: A Case Study on Acupuncture Point Locations

arXiv

引用

arXiv 2024年

作者： Li, Yiming Peng, Xueqing Li, Jianfu Zuo, Xu Peng, Suyuan Pei, Donghong Tao, Cui Xu, Hua Hong, Na McWilliams School of Biomedical Informatics The University of Texas Health Science Center at Houston HoustonTX77030 United States Section of Biomedical Informatics and Data Science School of Medicine Yale University New HavenCT06510 United States Department of Artificial Intelligence and Informatics Mayo Clinic JacksonvilleFL32224 United States Institute of Information on Traditional Chinese Medicine China Academy of Chinese Medical Sciences Beijing100010 China The University of Texas MD Anderson Cancer Center HoustonTX77030 United States

In acupuncture therapy, the accurate location of acupoints is essential for its effectiveness. The advanced language understanding capabilities of large language models (LLMs) like Generative Pre-trained Transformers (GPT) present a significant opportunity for extracting relations related to acupoint locations from textual knowledge sources. This study aims to compare the performance of GPT with traditional deep learning models (Long Short-Term Memory (LSTM) and Bidirectional Encoder Representations from Transformers for biomedical Text Mining (BioBERT)) in extracting acupoint-related location relations and assess the impact of pretraining and fine-tuning on GPT's performance. Materials and Methods We utilized the World Health Organization Standard Acupuncture Point Locations in the Western Pacific Region (WHO Standard) as our corpus, which consists of descriptions of 361 acupoints. Five types of relations ('direction_of,' 'distance_of,' 'part_of,' 'near_acupoint,' and 'located_near') (n= 3,174) between acupoints were annotated. Five models were compared: BioBERT, LSTM, pre-trained GPT-3.5, fine-tuned GPT-3.5, as well as pre-trained GPT-4. Performance metrics included micro-average exact match precision, recall, and F1 scores. Results Our results demonstrate that fine-tuned GPT-3.5 consistently outperformed other models in F1 scores across all relation types. Overall, it achieved the highest micro-average F1 score of 0.92. Conclusion This study underscores the effectiveness of LLMs like GPT in extracting relations related to acupoint locations, with implications for accurately modeling acupuncture knowledge and promoting standard implementation in acupuncture training and practice. The findings also contribute to advancing informatics applications in traditional and complementary medicine, showcasing the potential of LLMs in natural language processing. Copyright © 2024, The Authors. All rights reserved.

关键词： Acupuncture

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：