Abstract:This study investigates uncertainty quantification in large language models (LLMs) for medical applications, emphasizing both technical innovations and philosophical implications. As LLMs become integral to clinical decision-making, accurately communicating uncertainty is crucial for ensuring reliable, safe, and ethical AI-assisted healthcare. Our research frames uncertainty not as a barrier but as an essential part of knowledge that invites a dynamic and reflective approach to AI design. By integrating advanced probabilistic methods such as Bayesian inference, deep ensembles, and Monte Carlo dropout with linguistic analysis that computes predictive and semantic entropy, we propose a comprehensive framework that manages both epistemic and aleatoric uncertainties. The framework incorporates surrogate modeling to address limitations of proprietary APIs, multi-source data integration for better context, and dynamic calibration via continual and meta-learning. Explainability is embedded through uncertainty maps and confidence metrics to support user trust and clinical interpretability. Our approach supports transparent and ethical decision-making aligned with Responsible and Reflective AI principles. Philosophically, we advocate accepting controlled ambiguity instead of striving for absolute predictability, recognizing the inherent provisionality of medical knowledge.
Abstract:Introduction: This study provides a comprehensive performance assessment of vision-language models (VLMs) against established convolutional neural networks (CNNs) and classic machine learning models (CMLs) for computer-aided detection (CADe) and computer-aided diagnosis (CADx) of colonoscopy polyp images. Method: We analyzed 2,258 colonoscopy images with corresponding pathology reports from 428 patients. We preprocessed all images using standardized techniques (resizing, normalization, and augmentation) and implemented a rigorous comparative framework evaluating 11 distinct models: ResNet50, 4 CMLs (random forest, support vector machine, logistic regression, decision tree), two specialized contrastive vision language encoders (CLIP, BiomedCLIP), and three general-purpose VLMs ( GPT-4 Gemini-1.5-Pro, Claude-3-Opus). Our performance assessment focused on two clinical tasks: polyp detection (CADe) and classification (CADx). Result: In polyp detection, ResNet50 achieved the best performance (F1: 91.35%, AUROC: 0.98), followed by BiomedCLIP (F1: 88.68%, AUROC: [AS1] ). GPT-4 demonstrated comparable effectiveness to traditional machine learning approaches (F1: 81.02%, AUROC: [AS2] ), outperforming other general-purpose VLMs. For polyp classification, performance rankings remained consistent but with lower overall metrics. ResNet50 maintained the highest efficacy (weighted F1: 74.94%), while GPT-4 demonstrated moderate capability (weighted F1: 41.18%), significantly exceeding other VLMs (Claude-3-Opus weighted F1: 25.54%, Gemini 1.5 Pro weighted F1: 6.17%). Conclusion: CNNs remain superior for both CADx and CADe tasks. However, VLMs like BioMedCLIP and GPT-4 may be useful for polyp detection tasks where training CNNs is not feasible.
Abstract:This study evaluated self-reported response certainty across several large language models (GPT, Claude, Llama, Phi, Mistral, Gemini, Gemma, and Qwen) using 300 gastroenterology board-style questions. The highest-performing models (GPT-o1 preview, GPT-4o, and Claude-3.5-Sonnet) achieved Brier scores of 0.15-0.2 and AUROC of 0.6. Although newer models demonstrated improved performance, all exhibited a consistent tendency towards overconfidence. Uncertainty estimation presents a significant challenge to the safe use of LLMs in healthcare. Keywords: Large Language Models; Confidence Elicitation; Artificial Intelligence; Gastroenterology; Uncertainty Quantification
Abstract:Background: This study aimed to evaluate and compare the performance of classical machine learning models (CMLs) and large language models (LLMs) in predicting mortality associated with COVID-19 by utilizing a high-dimensional tabular dataset. Materials and Methods: We analyzed data from 9,134 COVID-19 patients collected across four hospitals. Seven CML models, including XGBoost and random forest (RF), were trained and evaluated. The structured data was converted into text for zero-shot classification by eight LLMs, including GPT-4 and Mistral-7b. Additionally, Mistral-7b was fine-tuned using the QLoRA approach to enhance its predictive capabilities. Results: Among the CML models, XGBoost and RF achieved the highest accuracy, with F1 scores of 0.87 for internal validation and 0.83 for external validation. In the LLM category, GPT-4 was the top performer with an F1 score of 0.43. Fine-tuning Mistral-7b significantly improved its recall from 1% to 79%, resulting in an F1 score of 0.74, which was stable during external validation. Conclusion: While LLMs show moderate performance in zero-shot classification, fine-tuning can significantly enhance their effectiveness, potentially aligning them closer to CML models. However, CMLs still outperform LLMs in high-dimensional tabular data tasks.
Abstract:Large language models (LLMs) hold great promise in summarizing medical evidence. Most recent studies focus on the application of proprietary LLMs. Using proprietary LLMs introduces multiple risk factors, including a lack of transparency and vendor dependency. While open-source LLMs allow better transparency and customization, their performance falls short compared to proprietary ones. In this study, we investigated to what extent fine-tuning open-source LLMs can further improve their performance in summarizing medical evidence. Utilizing a benchmark dataset, MedReview, consisting of 8,161 pairs of systematic reviews and summaries, we fine-tuned three broadly-used, open-sourced LLMs, namely PRIMERA, LongT5, and Llama-2. Overall, the fine-tuned LLMs obtained an increase of 9.89 in ROUGE-L (95% confidence interval: 8.94-10.81), 13.21 in METEOR score (95% confidence interval: 12.05-14.37), and 15.82 in CHRF score (95% confidence interval: 13.89-16.44). The performance of fine-tuned LongT5 is close to GPT-3.5 with zero-shot settings. Furthermore, smaller fine-tuned models sometimes even demonstrated superior performance compared to larger zero-shot models. The above trends of improvement were also manifested in both human and GPT4-simulated evaluations. Our results can be applied to guide model selection for tasks demanding particular domain knowledge, such as medical evidence summarization.
Abstract:Background: Evidence-based medicine (EBM) is fundamental to modern clinical practice, requiring clinicians to continually update their knowledge and apply the best clinical evidence in patient care. The practice of EBM faces challenges due to rapid advancements in medical research, leading to information overload for clinicians. The integration of artificial intelligence (AI), specifically Generative Large Language Models (LLMs), offers a promising solution towards managing this complexity. Methods: This study involved the curation of real-world clinical cases across various specialties, converting them into .json files for analysis. LLMs, including proprietary models like ChatGPT 3.5 and 4, Gemini Pro, and open-source models like LLaMA v2 and Mixtral-8x7B, were employed. These models were equipped with tools to retrieve information from case files and make clinical decisions similar to how clinicians must operate in the real world. Model performance was evaluated based on correctness of final answer, judicious use of tools, conformity to guidelines, and resistance to hallucinations. Results: GPT-4 was most capable of autonomous operation in a clinical setting, being generally more effective in ordering relevant investigations and conforming to clinical guidelines. Limitations were observed in terms of model ability to handle complex guidelines and diagnostic nuances. Retrieval Augmented Generation made recommendations more tailored to patients and healthcare systems. Conclusions: LLMs can be made to function as autonomous practitioners of evidence-based medicine. Their ability to utilize tooling can be harnessed to interact with the infrastructure of a real-world healthcare system and perform the tasks of patient management in a guideline directed manner. Prompt engineering may help to further enhance this potential and transform healthcare for the clinician and the patient.
Abstract:Diagnostic codes for Barrett's esophagus (BE), a precursor to esophageal cancer, lack granularity and precision for many research or clinical use cases. Laborious manual chart review is required to extract key diagnostic phenotypes from BE pathology reports. We developed a generalizable transformer-based method to automate data extraction. Using pathology reports from Columbia University Irving Medical Center with gastroenterologist-annotated targets, we performed binary dysplasia classification as well as granularized multi-class BE-related diagnosis classification. We utilized two clinically pre-trained large language models, with best model performance comparable to a highly tailored rule-based system developed using the same data. Binary dysplasia extraction achieves 0.964 F1-score, while the multi-class model achieves 0.911 F1-score. Our method is generalizable and faster to implement as compared to a tailored rule-based approach.