上部消化管病理におけるAIの将来像—食道癌・胃癌のAI病理診断実用化に向けての展望

Perspectives on the Clinical Implementation of AI-based Pathological Diagnosis for Esophageal and Gastric Cancer

要旨●近年，人工知能（AI）を用いた病理診断支援技術が注目されているが，実臨床への導入には多くの課題が残されている．特に食道癌・胃癌に限った話ではないが，病理診断は，画像だけでなく臨床情報や文脈に基づく複合的判断を要し，AIによる代替は容易ではない．また，病理AIは内視鏡AIと異なり，アナログ画像のデジタル化や正解ラベルのあいまいさ，ドメインシフトへの対応といった独自の困難を抱える．本稿では，病理AIの技術的発展と実装に向けた現状と課題を概説し，医師とAIの協働による未来像を展望する．

　In recent years, artificial intelligence（AI）-based technologies that support pathological diagnosis have received significant attention. However, several challenges persist before implementing such systems in clinical practice. Although not limited to esophageal and gastric cancers, pathological diagnosis requires complex decision-making according to images, clinical information, and contextual understanding. These are factors that make AI substitution challenging. Unlike endoscopic AI, pathology AI faces unique difficulties, including analog-to-digital image conversion requirements, ground truth label ambiguities, and domain shift issues. This study describes the progress and challenges in pathology AI and discusses the prospects of collaboration between physicians and AI.