Using Large Language Models for Medical Diagnosis: A Survey and Future Trends

Abstract

The rapid advancement of Large Language Models (LLMs) and Multimodal Large Language Models (MLLMs) has opened new possibilities for AI-driven medical diagnostics, offering enhanced decision-making capabilities and improved patient care. This paper provides a comprehensive review of research utilizing LLMs and MLLMs in healthcare, with a specific focus on medical diagnosis. We conducted a systematic literature review following a structured three-phase approach: Identification, Screening, and Inclusion. Through an extensive search on Google Scholar, we initially identified 293 articles, which were then refined through rigorous screening and eligibility criteria, ultimately selecting 19 key studies that directly applied LLMs and MLLMs in medical diagnosis. Our findings reveal that state-of-the-art LLMs, such as GPT-4, LLaMA3, and LLaVA-1.5, demonstrate strong performance in various medical domains, including radiology, dermatology, ophthalmology, and emotion recognition. These models facilitate accurate disease detection, multimodal analysis, and real-time diagnostic assistance. However, challenges remain in areas such as interpretability, bias mitigation, and regulatory compliance, necessitating further research and validation before widespread clinical adoption. Future trends indicate that LLMs will be increasingly integrated into clinical workflows, real-time decision-support systems, and telemedicine platforms. Advances in explainable AI, federated learning, and privacy-preserving AI will be crucial for ensuring trustworthy and ethical deployment in healthcare. This study highlights the transformative potential of LLMs in reshaping medical diagnostics while emphasizing the importance of responsible AI development and regulatory oversight to ensure safe and equitable implementation.