Impact of Large Language Model–assisted Differential Diagnosis on Clinical Decision-making in Dermatology: A Feasibility Study Using ChatGPT-5

Abstract

Recent advances in artificial intelligence (AI) have led to substantial progress in diagnostic support research in dermatology.In particular, deep learning-based image analysis has achieved dermatologist-level performance for various skin diseases, including cutaneous tumours, and has been shown to improve diagnostic accuracy among less experienced clinicians (1-3).More recently, large language models (LLMs) have attracted attention as diagnostic support tools capable of generating clinically relevant differential diagnoses from textual case descriptions(4).In our previous study, ChatGPT-4o demonstrated diagnostic accuracy comparable to that of board-certified dermatologists when provided with structured dermatological case information (5), suggesting its potential utility as a supportive aid in clinical decision-making.Despite these advances, how junior dermatology residents interact with AI-generated diagnostic suggestions in real-world settings remains insufficiently understood.Trainee physicians may either over-rely on or underutilize AI assistance, and incorrect AI outputs may negatively influence clinical decisions (6-8).We therefore conducted a feasibility study to evaluate the impact of LLM-assisted differential diagnosis on diagnostic accuracy among junior dermatology residents.

Ähnliche Arbeiten