AMIA Joint Summits on Translational Science proceedings. AMIA Joint Summits on Translational Science最新文献

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.

Monitoring the health of individuals during physically demanding tasks, such as the Hajj pilgrimage, requires robust methods for real-time detection of health-relevant contexts, including physical tiredness, emotional mood, and activity type. This paper introduces a positionally encoded Transformer model designed to detect these contexts from time-series data collected via wearable sensors. The model leverages Long Short-Term Memory (LSTM) for feature extraction and Transformer layers for context classification, utilizing positional encoding to capture the sequential dependencies within the sensor data. Our experiments, using data from 19 participants, show that the proposed model achieves high classification accuracy across multiple health-relevant contexts, significantly improving real-time health monitoring.

The Systematic reviews (SRs) for Idiopathic Pulmonary Fibrosis (IPF) play a crucial role in guiding evidence-based healthcare by synthesizing data across multiple studies. A key factor in ensuring the reliability and applicability of these reviews is the geographic diversity of the authors involved, as this can significantly influence the generalizability of findings. Traditional 2D maps used to visualize author locations often fall short in capturing the depth and regional disparities effectively, as overlapping points or dense clusters can obscure critical details, resulting in an incomplete view of geographic distribution. To address these limitations, this study introduces a novel approach that combines a 3D geographic map and a Temporal-Spatial Graph Attention Network (TS-GAT) to assess and visualize the geographic diversity of authors in SRs on IPF. The 3D visualization provides an enhanced, layered representation of author locations, revealing hidden regional disparities and biases. The TS-GAT captures both temporal and spatial relationships in the author collaboration network, allowing for deeper insights into the evolution of geographic representation over time. This integrated approach aims to uncover potential biases in global representation, offering a comprehensive understanding of the geographic spread and temporal trends in authorship within SRs, ultimately contributing to more balanced and inclusive evidence synthesis.

Automated text summarization (ATS) is crucial for collecting specialized, domain-specific information. Zero-shot learning (ZSL) allows large language models (LLMs) to respond to prompts on information not included in their training, playing a vital role in this process. This study evaluates LLMs' effectiveness in generating accurate summaries under ZSL conditions and explores using retrieval augmented generation (RAG) and prompt engineering to enhance factual accuracy and understanding. We combined LLMs with summarization modeling, prompt engineering, and RAG, evaluating the summaries using the METEOR metric and keyword frequencies through word clouds. Results indicate that LLMs are generally well-suited for ATS tasks, demonstrating an ability to handle specialized information under ZSL conditions with RAG. However, web scraping limitations hinder a single generalized retrieval mechanism. While LLMs show promise for ATS under ZSL conditions with RAG, challenges like goal misgeneralization and web scraping limitations need addressing. Future research should focus on solutions to these issues.

Large language models (LLMs) such as ChatGPT are fine-tuned on large and diverse instruction-following corpora, and can generalize to new tasks. However, those instruction-tuned LLMs often perform poorly in specialized medical natural language understanding (NLU) tasks that require domain knowledge, granular text comprehension, and structured data extraction. To bridge the gap, we: (1) propose a unified prompting format for 7 important NLU tasks, (2) curate an instruction-tuning dataset, MNLU-Instruct, utilizing diverse existing open-source medical NLU corpora, and (3) develop BioMistral-NLU, a generalizable medical NLU model, through fine-tuning BioMistral on MNLU-Instruct. We evaluate BioMistral-NLU in a zero-shot setting, across 6 important NLU tasks, from two widely adopted medical NLU benchmarks: BLUE and BLURB. Our experiments show that our BioMistral-NLU outperforms the original BioMistral, as well as the proprietary LLMs - ChatGPT and GPT-4. Our dataset-agnostic prompting strategy and instruction tuning step over diverse NLU tasks enhance LLMs' generalizability across diverse medical NLU tasks. Our ablation experiments show that instruction-tuning on a wider variety of tasks, even when the total number of training instances remains constant, enhances downstream zero-shot generalization.

Chronic Kidney Disease (CKD) is a significant global public health issue, affecting over 10% of the population. Timely diagnosis is crucial for effective management. Leveraging machine learning within healthcare offers promising advancements in predictive diagnostics. We developed a Web-Based Clinical Decision Support System (CDSS) for CKD, incorporating advanced Explainable AI (XAI) methods, specifically SHAP (Shapley Additive Explanations) and LIME (Local Interpretable Model-agnostic Explanations). The model employs and evaluates multiple classifiers: KNN, Random Forest, AdaBoost, XGBoost, CatBoost, and Extra Trees, to predict CKD. The effectiveness of the models is assessed by measuring their accuracy, analyzing confusion matrix statistics, and the AUC. AdaBoost achieved a 100% accuracy rate. Except for KNN, all classifiers consistently reached perfect precision and sensitivity. Additionally, we present a real-time web-based application to operationalize the model, enhancing trust and accessibility for healthcare practitioners and stakeholder.

Idiopathic pulmonary fibrosis (IPF) is a rare disease that is challenging to diagnose. Patients with IPF often spend years awaiting a diagnosis after the onset of initial respiratory symptoms, and only a small percentage receive antifibrotic treatment. In this study, we examine the associations between social determinants of health (SDoH) and two critical factors: time to IPF diagnosis following the onset of initial respiratory symptoms, and whether the patient receives antifibrotic treatment. To approximate individual SDoH characteristics, we extract demographic-specific averages from zip code-level data using the American Community Survey (via the U.S. Census Bureau API). Two classification models are constructed, including logistic regression and XGBoost classification. The results indicate that for time-to-diagnosis, the top three SDoH factors are education, gender, and insurance coverage. Patients with higher education levels and better insurance are more likely to receive a quicker diagnosis, with males having an advantage over females. For antifibrotic treatment, the top three SDoH factors are insurance, gender, and race. Patients with better insurance coverage are more likely to receive antifibrotic treatment, with males and White patients having an advantage over females and patients of other ethnicities. This research may help address disparities in the diagnosis and treatment of IPF related to socioeconomic status.

Sudden Unexpected Death in Epilepsy (SUDEP) is a major cause of death for epilepsy patients having uncontrolled seizures. Understanding the complex neural circuits within the central nervous system is crucial for understanding the mechanisms underlying cardiorespiratory regulation, particularly in the context of SUDEP. This study explores the potential of GPT-4o, a cutting-edge language model, to automate the extraction of neural projections from scientific literature. We developed prompts to extract neuroscientific structures, extract projections, and perform synonym harmonization. Applying the approach to four neuroscientific articles, the method extracted 205 projections. A random sample of 100 projections identified was handed over to a domain expert for review where 95 were found to be correct. Therefore, GPT-4o was determined to be accurate in parsing complex scientific texts in extracting neural projections. Future work will involve extracting additional entities like techniques and species information for the projections identified.

Growing evidence suggests that social determinants of health (SDoH), a set of nonmedical factors, affect individuals' risks of developing Alzheimer's disease (AD) and related dementias. Nevertheless, the etiological mechanisms underlying such relationships remain largely unclear, mainly due to difficulties in collecting relevant information. This study presents a novel, automated framework that leverages recent advancements of large language model (LLM) and natural language processing techniques to mine SDoH knowledge from extensive literature and integrate it with AD-related biological entities extracted from the general-purpose knowledge graph PrimeKG. Utilizing graph neural networks, we performed link prediction tasks to evaluate the resultant SDoH-augmented knowledge graph. Our framework shows promise for enhancing knowledge discovery in AD and can be generalized to other SDoH-related research areas, offering a new tool for exploring the impact of social determinants on health outcomes. Our code is available at: https://github.com/hwq0726/SDoHenPKG.

Diagnosis prediction is a critical task in healthcare, where timely and accurate identification of medical conditions can significantly impact patient outcomes. Traditional machine learning and deep learning models have achieved notable success in this domain but often lack interpretability which is a crucial requirement in clinical settings. In this study, we explore the use of neuro-symbolic methods, specifically Logical Neural Networks (LNNs), to develop explainable models for diagnosis prediction. Essentially, we design and implement LNN-based models that integrate domain-specific knowledge through logical rules with learnable weights and thresholds. Our models, particularly Mmulti-pathway and Mcomprehensive, demonstrate superior performance over traditional models such as Logistic Regression, SVM, and Random Forest, achieving higher accuracy (up to 80.52%) and AUROC scores (up to 0.8457) in the case study of diabetes prediction. The learned weights and thresholds within the LNN models provide direct insights into feature contributions, enhancing interpretability without compromising predictive power. These findings highlight the potential of neuro-symbolic approaches in bridging the gap between accuracy and explainability in healthcare AI applications. By offering transparent and adaptable diagnostic models, our work contributes to the advancement ofprecision medicine and supports the development of equitable healthcare solutions. Future research will focus on extending these methods to larger and more diverse datasets to further validate their applicability across different medical conditions and populations.