Timely disease diagnosis is challenging due to increasing disease burdens and limited clinician availability. AI shows promise in diagnosis accuracy but faces real-world application issues due to insufficient validation in clinical workflows and diverse populations. This study addresses gaps in medical AI downstream accountability through a case study on age-related macular degeneration (AMD) diagnosis and severity classification. We designed and implemented an AI-assisted diagnostic workflow for AMD, comparing diagnostic performance with and without AI assistance among 24 clinicians from 12 institutions with real patient data sampled from the Age-Related Eye Disease Study (AREDS). Additionally, we demonstrated continual enhancement of an existing AI model by incorporating approximately 40,000 additional medical images (named AREDS2 dataset). The improved model was then systematically evaluated using both AREDS and AREDS2 test sets, as well as an external test set from Singapore. AI assistance markedly enhanced diagnostic accuracy and classification for 23 out of 24 clinicians, with the average F1-score increasing by 20% from 37.71 (Manual) to 45.52 (Manual + AI) (P-value < 0.0001), achieving an improvement of over 50% in some cases. In terms of efficiency, AI assistance reduced diagnostic times for 17 out of the 19 clinicians tracked, with time savings of up to 40%. Furthermore, a model equipped with continual learning showed robust performance across three independent datasets, recording a 29% increase in accuracy, and elevating the F1-score from 42 to 54 in the Singapore population.

Despite the potential of Large Language Models (LLMs) in medicine, they may generate responses lacking supporting evidence or based on hallucinated evidence. While Retrieval Augment Generation (RAG) is popular to address this issue, few studies implemented and evaluated RAG in downstream domain-specific applications. We developed a RAG pipeline with 70,000 ophthalmology-specific documents that retrieve relevant documents to augment LLMs during inference time. In a case study on long-form consumer health questions, we systematically evaluated the responses including over 500 references of LLMs with and without RAG on 100 questions with 10 healthcare professionals. The evaluation focuses on factuality of evidence, selection and ranking of evidence, attribution of evidence, and answer accuracy and completeness. LLMs without RAG provided 252 references in total. Of which, 45.3% hallucinated, 34.1% consisted of minor errors, and 20.6% were correct. In contrast, LLMs with RAG significantly improved accuracy (54.5% being correct) and reduced error rates (18.8% with minor hallucinations and 26.7% with errors). 62.5% of the top 10 documents retrieved by RAG were selected as the top references in the LLM response, with an average ranking of 4.9. The use of RAG also improved evidence attribution (increasing from 1.85 to 2.49 on a 5-point scale, P<0.001), albeit with slight decreases in accuracy (from 3.52 to 3.23, P=0.03) and completeness (from 3.47 to 3.27, P=0.17). The results demonstrate that LLMs frequently exhibited hallucinated and erroneous evidence in the responses, raising concerns for downstream applications in the medical domain. RAG substantially reduced the proportion of such evidence but encountered challenges.

As opposed to evaluating computation and logic-based reasoning, current benchmarks for evaluating large language models (LLMs) in medicine are primarily focused on question-answering involving domain knowledge and descriptive reasoning. While such qualitative capabilities are vital to medical diagnosis, in real-world scenarios, doctors frequently use clinical calculators that follow quantitative equations and rule-based reasoning paradigms for evidence-based decision support. To this end, we propose MedCalc-Bench, a first-of-its-kind dataset focused on evaluating the medical calculation capability of LLMs. MedCalc-Bench contains an evaluation set of over 1000 manually reviewed instances from 55 different medical calculation tasks. Each instance in MedCalc-Bench consists of a patient note, a question requesting to compute a specific medical value, a ground truth answer, and a step-by-step explanation showing how the answer is obtained. While our evaluation results show the potential of LLMs in this area, none of them are effective enough for clinical settings. Common issues include extracting the incorrect entities, not using the correct equation or rules for a calculation task, or incorrectly performing the arithmetic for the computation. We hope our study highlights the quantitative knowledge and reasoning gaps in LLMs within medical settings, encouraging future improvements of LLMs for various clinical calculation tasks.

Automatic medical discovery by AI is a dream of many. One step toward that goal is to create an AI model to understand clinical studies and synthesize clinical evidence from the literature. Clinical evidence synthesis currently relies on systematic reviews of clinical trials and retrospective analyses from medical literature. However, the rapid expansion of publications presents challenges in efficiently identifying, summarizing, and updating evidence. We introduce TrialMind, a generative AI-based pipeline for conducting medical systematic reviews, encompassing study search, screening, and data extraction phases. We utilize large language models (LLMs) to drive each pipeline component while incorporating human expert oversight to minimize errors. To facilitate evaluation, we also create a benchmark dataset TrialReviewBench, a custom dataset with 870 annotated clinical studies from 25 meta-analysis papers across various medical treatments. Our results demonstrate that TrialMind significantly improves the literature review process, achieving high recall rates (0.897-1.000) in study searching from over 20 million PubMed studies and outperforming traditional language model embeddings-based methods in screening (Recall@20 of 0.227-0.246 vs. 0.000-0.102). Furthermore, our approach surpasses direct GPT-4 performance in result extraction, with accuracy ranging from 0.65 to 0.84. We also support clinical evidence synthesis in forest plots, as validated by eight human annotators who preferred TrialMind over the GPT-4 baseline with a winning rate of 62.5%-100% across the involved reviews. Our findings suggest that an LLM-based clinical evidence synthesis approach, such as TrialMind, can enable reliable and high-quality clinical evidence synthesis to improve clinical research efficiency.

Automatically interpreting CT scans can ease the workload of radiologists. However, this is challenging mainly due to the scarcity of adequate datasets and reference standards for evaluation. This study aims to bridge this gap by introducing a novel evaluation framework, named ``GPTRadScore''. This framework assesses the capabilities of multi-modal LLMs, such as GPT-4 with Vision (GPT-4V), Gemini Pro Vision, LLaVA-Med, and RadFM, in generating descriptions for prospectively-identified findings. By employing a decomposition technique based on GPT-4, GPTRadScore compares these generated descriptions with gold-standard report sentences, analyzing their accuracy in terms of body part, location, and type of finding. Evaluations demonstrated a high correlation with clinician assessments and highlighted its potential over traditional metrics, such as BLEU, METEOR, and ROUGE. Furthermore, to contribute to future studies, we plan to release a benchmark dataset annotated by clinicians. Using GPTRadScore, we found that while GPT-4V and Gemini Pro Vision fare better, their performance revealed significant areas for improvement, primarily due to limitations in the dataset used for training these models. To demonstrate this potential, RadFM was fine-tuned and it resulted in significant accuracy improvements: location accuracy rose from 3.41\% to 12.8\%, body part accuracy from 29.12\% to 53\%, and type accuracy from 9.24\% to 30\%, thereby validating our hypothesis.

Training a neural network-based biomedical named entity recognition (BioNER) model usually requires extensive and costly human annotations. While several studies have employed multi-task learning with multiple BioNER datasets to reduce human effort, this approach does not consistently yield performance improvements and may introduce label ambiguity in different biomedical corpora. We aim to tackle those challenges through transfer learning from easily accessible resources with fewer concept overlaps with biomedical datasets. In this paper, we proposed GERBERA, a simple-yet-effective method that utilized a general-domain NER dataset for training. Specifically, we performed multi-task learning to train a pre-trained biomedical language model with both the target BioNER dataset and the general-domain dataset. Subsequently, we fine-tuned the models specifically for the BioNER dataset. We systematically evaluated GERBERA on five datasets of eight entity types, collectively consisting of 81,410 instances. Despite using fewer biomedical resources, our models demonstrated superior performance compared to baseline models trained with multiple additional BioNER datasets. Specifically, our models consistently outperformed the baselines in six out of eight entity types, achieving an average improvement of 0.9% over the best baseline performance across eight biomedical entity types sourced from five different corpora. Our method was especially effective in amplifying performance on BioNER datasets characterized by limited data, with a 4.7% improvement in F1 scores on the JNLPBA-RNA dataset.

The integration of Large Language Models (LLMs) into healthcare applications offers promising advancements in medical diagnostics, treatment recommendations, and patient care. However, the susceptibility of LLMs to adversarial attacks poses a significant threat, potentially leading to harmful outcomes in delicate medical contexts. This study investigates the vulnerability of LLMs to two types of adversarial attacks in three medical tasks. Utilizing real-world patient data, we demonstrate that both open-source and proprietary LLMs are susceptible to manipulation across multiple tasks. This research further reveals that domain-specific tasks demand more adversarial data in model fine-tuning than general domain tasks for effective attack execution, especially for more capable models. We discover that while integrating adversarial data does not markedly degrade overall model performance on medical benchmarks, it does lead to noticeable shifts in fine-tuned model weights, suggesting a potential pathway for detecting and countering model attacks. This research highlights the urgent need for robust security measures and the development of defensive mechanisms to safeguard LLMs in medical applications, to ensure their safe and effective deployment in healthcare settings.

Gene set knowledge discovery is essential for advancing human functional genomics. Recent studies have shown promising performance by harnessing the power of Large Language Models (LLMs) on this task. Nonetheless, their results are subject to several limitations common in LLMs such as hallucinations. In response, we present GeneAgent, a first-of-its-kind language agent featuring self-verification capability. It autonomously interacts with various biological databases and leverages relevant domain knowledge to improve accuracy and reduce hallucination occurrences. Benchmarking on 1,106 gene sets from different sources, GeneAgent consistently outperforms standard GPT-4 by a significant margin. Moreover, a detailed manual review confirms the effectiveness of the self-verification module in minimizing hallucinations and generating more reliable analytical narratives. To demonstrate its practical utility, we apply GeneAgent to seven novel gene sets derived from mouse B2905 melanoma cell lines, with expert evaluations showing that GeneAgent offers novel insights into gene functions and subsequently expedites knowledge discovery.

Deep learning has enabled breakthroughs in automated diagnosis from medical imaging, with many successful applications in ophthalmology. However, standard medical image classification approaches only assess disease presence at the time of acquisition, neglecting the common clinical setting of longitudinal imaging. For slow, progressive eye diseases like age-related macular degeneration (AMD) and primary open-angle glaucoma (POAG), patients undergo repeated imaging over time to track disease progression and forecasting the future risk of developing disease is critical to properly plan treatment. Our proposed Longitudinal Transformer for Survival Analysis (LTSA) enables dynamic disease prognosis from longitudinal medical imaging, modeling the time to disease from sequences of fundus photography images captured over long, irregular time periods. Using longitudinal imaging data from the Age-Related Eye Disease Study (AREDS) and Ocular Hypertension Treatment Study (OHTS), LTSA significantly outperformed a single-image baseline in 19/20 head-to-head comparisons on late AMD prognosis and 18/20 comparisons on POAG prognosis. A temporal attention analysis also suggested that, while the most recent image is typically the most influential, prior imaging still provides additional prognostic value.

Expert curation is essential to capture knowledge of enzyme functions from the scientific literature in FAIR open knowledgebases but cannot keep pace with the rate of new discoveries and new publications. In this work we present EnzChemRED, for Enzyme Chemistry Relation Extraction Dataset, a new training and benchmarking dataset to support the development of Natural Language Processing (NLP) methods such as (large) language models that can assist enzyme curation. EnzChemRED consists of 1,210 expert curated PubMed abstracts in which enzymes and the chemical reactions they catalyze are annotated using identifiers from the UniProt Knowledgebase (UniProtKB) and the ontology of Chemical Entities of Biological Interest (ChEBI). We show that fine-tuning pre-trained language models with EnzChemRED can significantly boost their ability to identify mentions of proteins and chemicals in text (Named Entity Recognition, or NER) and to extract the chemical conversions in which they participate (Relation Extraction, or RE), with average F1 score of 86.30% for NER, 86.66% for RE for chemical conversion pairs, and 83.79% for RE for chemical conversion pairs and linked enzymes. We combine the best performing methods after fine-tuning using EnzChemRED to create an end-to-end pipeline for knowledge extraction from text and apply this to abstracts at PubMed scale to create a draft map of enzyme functions in literature to guide curation efforts in UniProtKB and the reaction knowledgebase Rhea. The EnzChemRED corpus is freely available at https://ftp.expasy.org/databases/rhea/nlp/.