Evaluating the Limitations of Large Language Models in Medical Software Systems Engineering

Large Language Models (LLMs) have demonstrated impressive advancements in a range of technological fields. However, their application to medical software systems engineering presents significant challenges. This paper systematically investigates the limitations of LLMs in this context, focusing on the critical nature of medical systems and the stringent regulatory requirements that constrain their use. Key limitations identified include issues with output accuracy, model interpretability, compliance with medical standards such as IEC ۶۲۳۰۴ and ISO ۱۳۴۸۵, security concerns when handling sensitive medical data, and ethical ambiguities regarding accountability. Through a detailed case study analysis and an extensive review of existing literature, the paper highlights how these challenges can negatively affect the medical software development lifecycle. The findings suggest that the direct implementation of LLMs in medical systems, without proper safeguards, could pose serious risks to patient safety and system reliability. The paper proposes several practical solutions, including the development of specialized validation frameworks, combining rule-based and deep learning approaches, and designing hybrid human-machine architectures. In conclusion, the study outlines future research directions to address these challenges and advance the safe integration of AI technologies in healthcare.

کلیدواژه ها:

Large Language Models (LLMs) ، Medical Software Engineering ، AI in Healthcare ، Software Reliability ، Regulatory Compliance