HDTwin: Building a Human Digital Twin using Large Language Models for Cognitive Diagnosis (Preprint) DOI Creative Commons
Gina Sprint, Maureen Schmitter‐Edgecombe, Diane J. Cook

и другие.

JMIR Formative Research, Год журнала: 2024, Номер 8, С. e63866 - e63866

Опубликована: Ноя. 7, 2024

Background Human digital twins have the potential to change practice of personalizing cognitive health diagnosis because these systems can integrate multiple sources information and influence into a unified model. Cognitive is multifaceted, yet researchers clinical professionals struggle align diverse single Objective This study aims introduce method called HDTwin, for unifying heterogeneous data using large language models. HDTwin designed predict diagnoses offer explanations its inferences. Methods integrates from sources, including demographic, behavioral, ecological momentary assessment, n-back test, speech, baseline experimenter testing session markers. Data are converted text prompts The system then combines inputs with relevant external knowledge scientific literature construct predictive model’s performance validated 3 studies involving 124 participants, comparing diagnostic accuracy machine learning classifiers. Results achieves peak 0.81 based on automated selection markers, significantly outperforming On average, yielded accuracy=0.77, precision=0.88, recall=0.63, Matthews correlation coefficient=0.57. In comparison, classifiers average accuracy=0.65, precision=0.86, recall=0.35, coefficient=0.36. experiments also reveal that yields superior when fused compared sources. HDTwin’s chatbot interface provides interactive dialogues, aiding in interpretation allowing further exploration patient data. Conclusions data, enhancing explainability diagnoses. approach outperforms traditional models an navigating information. shows promise improving early detection intervention strategies health.

Язык: Английский

Large Language Models in Healthcare and Medical Domain: A Review DOI Creative Commons
Zabir Al Nazi, Wei Peng

Informatics, Год журнала: 2024, Номер 11(3), С. 57 - 57

Опубликована: Авг. 7, 2024

The deployment of large language models (LLMs) within the healthcare sector has sparked both enthusiasm and apprehension. These exhibit remarkable ability to provide proficient responses free-text queries, demonstrating a nuanced understanding professional medical knowledge. This comprehensive survey delves into functionalities existing LLMs designed for applications elucidates trajectory their development, starting with traditional Pretrained Language Models (PLMs) then moving present state in sector. First, we explore potential amplify efficiency effectiveness diverse applications, particularly focusing on clinical tasks. tasks encompass wide spectrum, ranging from named entity recognition relation extraction natural inference, multimodal document classification, question-answering. Additionally, conduct an extensive comparison most recent state-of-the-art domain, while also assessing utilization various open-source highlighting significance applications. Furthermore, essential performance metrics employed evaluate biomedical shedding light limitations. Finally, summarize prominent challenges constraints faced by offering holistic perspective benefits shortcomings. review provides exploration current landscape healthcare, addressing role transforming areas that warrant further research development.

Язык: Английский

Процитировано

72
Progress and trends in neurological disorders research based on deep learning DOI
Muhammad Shahid Iqbal, Md Belal Bin Heyat, Saba Parveen

и другие.

Computerized Medical Imaging and Graphics, Год журнала: 2024, Номер 116, С. 102400 - 102400

Опубликована: Май 25, 2024

Язык: Английский

Процитировано

9
Introduction to Large Language Models (LLMs) for dementia care and research DOI Creative Commons
Matthias S. Treder,

Sojin Lee,

Kamen A. Tsvetanov

и другие.

Frontiers in Dementia, Год журнала: 2024, Номер 3

Опубликована: Май 14, 2024

Dementia is a progressive neurodegenerative disorder that affects cognitive abilities including memory, reasoning, and communication skills, leading to gradual decline in daily activities social engagement. In light of the recent advent Large Language Models (LLMs) such as ChatGPT, this paper aims thoroughly analyse their potential applications usefulness dementia care research.

Язык: Английский

Процитировано

9
Pre-trained language models in medicine: A survey DOI Creative Commons
Xudong Luo,

Zhiqi Deng,

Binxia Yang

и другие.

Artificial Intelligence in Medicine, Год журнала: 2024, Номер 154, С. 102904 - 102904

Опубликована: Июнь 5, 2024

With the rapid progress in Natural Language Processing (NLP), Pre-trained Models (PLM) such as BERT, BioBERT, and ChatGPT have shown great potential various medical NLP tasks. This paper surveys cutting-edge achievements applying PLMs to Specifically, we first brief PLMS outline research of medicine. Next, categorise discuss types tasks NLP, covering text summarisation, question-answering, machine translation, sentiment analysis, named entity recognition, information extraction, education, relation mining. For each type task, provide an overview basic concepts, main methodologies, advantages PLMs, steps application, datasets for training testing, metrics task evaluation. Subsequently, a summary recent important findings is presented, analysing their motivations, strengths vs weaknesses, similarities differences, discussing limitations. Also, assess quality influence reviewed this by comparing citation count papers reputation impact conferences journals where they are published. Through these indicators, further identify most concerned topics currently. Finally, look forward future directions, including enhancing models' reliability, explainability, fairness, promote application clinical practice. In addition, survey also collect some download links model codes relevant datasets, which valuable references researchers techniques medicine professionals seeking enhance expertise healthcare service through AI technology.

Язык: Английский

Процитировано

8
Health Care Language Models and Their Fine-Tuning for Information Extraction: Scoping Review DOI Creative Commons
Miguel Nunes, João Boné, João C. Ferreira

и другие.

JMIR Medical Informatics, Год журнала: 2024, Номер 12, С. e60164 - e60164

Опубликована: Окт. 21, 2024

Background In response to the intricate language, specialized terminology outside everyday life, and frequent presence of abbreviations acronyms inherent in health care text data, domain adaptation techniques have emerged as crucial transformer-based models. This refinement knowledge language models (LMs) allows for a better understanding medical textual which results an improvement downstream tasks, such information extraction (IE). We identified gap literature regarding LMs. Therefore, this study presents scoping review investigating methods transformers care, differentiating between English non-English languages, focusing on Portuguese. Most specifically, we investigated development LMs, with aim comparing Portuguese other more developed languages guide path non–English-language fewer resources. Objective aimed research IE models, regardless understand efficacy what are entities most commonly extracted. Methods was conducted using PRISMA-ScR (Preferred Reporting Items Systematic reviews Meta-Analyses extension Scoping Reviews) methodology Scopus Web Science Core Collection databases. Only studies that mentioned creation LMs or were included, while large (LLMs) excluded. The latest not included since wanted LLMs, architecturally different distinct purposes. Results Our search query retrieved 137 studies, 60 met inclusion criteria, none them systematic reviews. Chinese developed. These already disease-specific others only general–health European does any public LM should take examples from develop, first, general-health then, advanced phase, Regarding used method, named entity recognition popular topic, few mentioning Assertion Status addressing lexical problems. extracted diagnosis, posology, symptoms. Conclusions findings indicate is beneficial, achieving tasks. analysis allowed us use languages. lacks relevant draw develop these drive progress AI. Health professionals could benefit highlighting medically optimizing reading be create patient timelines, allowing profiling.

Язык: Английский

Процитировано

5
Evaluating the validity of the nursing statements algorithmically generated based on the International Classifications of Nursing Practice for respiratory nursing care using large language models DOI
Hyeoneui Kim, Hyewon Park,

Sunghoon Kang

и другие.

Journal of the American Medical Informatics Association, Год журнала: 2024, Номер 31(6), С. 1397 - 1403

Опубликована: Апрель 17, 2024

Abstract Objective This study aims to facilitate the creation of quality standardized nursing statements in South Korea’s hospitals using algorithmic generation based on International Classifications Nursing Practice (ICNP) and evaluation through Large Language Models. Materials Methods We algorithmically generated 15 972 related acute respiratory care 117 concepts concept composition models ICNP. Human reviewers, Generative Pre-trained Transformers 4.0 (GPT-4.0), Bio_Clinical Bidirectional Encoder Representations from (BERT) evaluated for validity. The by GPT-4.0 Bio_ClinicalBERT was conducted with without contextual information training. Results Of statements, 2207 were deemed valid expert reviewers. showed a zero-shot AUC 0.857, which aggravated information. Bio_ClinicalBERT, after training, significantly improved, reaching an 0.998. Conclusion effectively validates auto-generated offering promising solution enhance streamline healthcare documentation processes.

Язык: Английский

Процитировано

4
The foundational capabilities of large language models in predicting postoperative risks using clinical notes DOI Creative Commons

Charles Alba,

Bing Xue, Joanna Abraham

и другие.

npj Digital Medicine, Год журнала: 2025, Номер 8(1)

Опубликована: Фев. 11, 2025

Clinical notes recorded during a patient's perioperative journey holds immense informational value. Advances in large language models (LLMs) offer opportunities for bridging this gap. Using 84,875 preoperative and its associated surgical cases from 2018 to 2021, we examine the performance of LLMs predicting six postoperative risks using various fine-tuning strategies. Pretrained outperformed traditional word embeddings by an absolute AUROC 38.3% AUPRC 33.2%. Self-supervised further improved 3.2% 1.5%. Incorporating labels into training increased 1.8% 2%. The highest was achieved with unified foundation model, improvements 3.6% 2.6% compared self-supervision, highlighting foundational capabilities risks, which could be potentially beneficial when deployed care.

Язык: Английский

Процитировано

0
Resistance training protects the hippocampus and precuneus against atrophy and benefits white matter integrity in older adults with mild cognitive impairment DOI
Isadora Cristina Ribeiro, Camila Vieira Ligo Teixeira, Thiago Junqueira Ribeiro de Rezende

и другие.

GeroScience, Год журнала: 2025, Номер unknown

Опубликована: Янв. 2, 2025

Язык: Английский

Процитировано

0
Multi-Modal Fusion of Routine Care Electronic Health Records (EHR): A Scoping Review DOI Creative Commons
Zina Ben Miled,

Jacob A. Shebesh,

Jing Su

и другие.

Information, Год журнала: 2025, Номер 16(1), С. 54 - 54

Опубликована: Янв. 15, 2025

Background: Electronic health records (EHR) are now widely available in healthcare institutions to document the medical history of patients as they interact with services. In particular, routine care EHR data collected for a large number patients.These span multiple heterogeneous elements (i.e., demographics, diagnosis, medications, clinical notes, vital signs, and laboratory results) which contain semantic, concept, temporal information. Recent advances generative learning techniques were able leverage fusion enhance decision support. Objective: A scoping review proposed including architectures, input elements, application areas is needed synthesize variances identify research gaps that can promote re-use these new outcomes. Design: comprehensive literature search was conducted using Google Scholar high impact architectures over multi-modal during period 2018 2023. The guidelines from PRISMA (Preferred Reporting Items Systematic Reviews Meta-Analyses) extension followed. findings derived selected studies thematic comparative analysis. Results: revealed lack standard definition transformed into modalities. These definitions ignore one or more key characteristics source, encoding scheme, concept level. Moreover, order adapt emergent techniques, classification should distinguish take consideration concurrently happen all three layers encoding, representation, decision). aspects constitute first step towards streamlined approach design data. addition, current pretrained models inconsistent their handling semantic information thereby hindering different applications settings. Conclusions: Current mostly follow design-by-example methodology. Guidelines efficient broad range applications. addition promoting re-use, need outline best practices combining modalities while leveraging transfer co-learning well encoding.

Язык: Английский

Процитировано

0
AI-powered model for accurate prediction of MCI-to-AD progression DOI Creative Commons
Ahmed Abdelhameed, Jingna Feng, Xinyue Hu

и другие.

Acta Pharmaceutica Sinica B, Год журнала: 2025, Номер unknown

Опубликована: Фев. 1, 2025

Язык: Английский

Процитировано

0