How Can Artificial Intelligence Alleviate Physician Overload: A Current Issue

There is a strong feeling that artificial intelligence (AI) is reshaping most human activities, and this is supported by impressive proof-of-concept reports that were published during the last decade. Furthermore, medicine is a prominent candidate for AI development, due to the overload on medical practitioners, and specific features of medical practice including the existence of general guidelines and diagnostic algorithms, development of evidence-based medicine and use of an accurate vocabulary making it highly suited for processing by artificial intelligence models. Thus, the current challenge is not to determine whether the introduction of AI in medicine is beneficial or even acceptable, but to organize this inevitable evolution to optimize benefit and prevent pitfalls. The aim of this paper is to discuss the main challenges that must be met to optimize AI support to medical practice. It is suggested that (i) specific procedures are needed to ensure the reliability, comprehensiveness and availability of medical datasets, (ii) there is a need for specific benchmarks and guidelines to validate AI use with a reliability comparable to that ensured by procedures currently used in medicine and (iii) the selection of an AI model to perform a given task must be examined to achieve an optimal balance between efficiency, feasibility and cost.