From the basics of medical image diagnosis using AI technology to generative AI

Abstract

  Artificial intelligence (AI) is evolving at a tremendous pace and is increasingly permeating the medical field. In the field of medical image diagnosis, the performance of so-called CAD (computer-aided diagnosis) systems that support doctors’ image diagnosis has significantly improved as AI-CAD due to the recent emergence of deep learning, the core technology of AI. AI products are being introduced into medical settings, with some even offering insurance assistance. Such AI is classified as “discriminative AI (recognition AI),” but with recent “generative AI,” AI-CAD will likely evolve into generative AI-CAD that is even closer to that of human doctors. This review provides an overview of the current state of AI utilization in medical care, explains the basics of deep learning and generative AI (large-scale language models, Transformer, ChatGPT, image generation AI, basic models, etc.), which are technologies that are attracting attention as a driving force, and then discusses the current status and challenges of introducing image diagnosis AI into clinical settings.