[Frontiers in Research, Development, and Clinical Application of AI-SaMDs]

Artificial intelligence (AI) technologies are rapidly being introduced into the medical device field, particularly in the form of software as a medical device (AI-SaMD), for which regulatory approvals are increasing. In the United States and Europe, many approved AI-SaMD products have been implemented in medical imaging applications, such as computed tomography (CT) and magnetic resonance imaging (MRI). Similar trends are observed in Japan, where approved AI-SaMD products have been implemented for endoscopy, laparoscopy, and CT. In gastrointestinal endoscopy, several randomized controlled trials have demonstrated that AI-assisted systems improve adenoma detection rate and reduce miss rate, contributing to overall procedural quality. In diagnostic imaging, AI has been used to develop techniques that reduce radiation dose while maintaining diagnostic performance, and has also been explored for use in screening examinations. In surgical assistance, the integration of AI with extended reality (XR) technologies has enabled intraoperative visualization of anatomical structures and instrument tracking, offering novel strategies for navigation. In pathology, deep learning models have shown potential to predict molecular subtypes and treatment responsiveness in malignancies. Collectively, these findings indicate that medical AI is transitioning from research-based technologies toward clinical implementation and, in some domains, commercialization. Future challenges include standardization, explainability, and effective integration into clinical workflows to ensure safe and sustainable adoption.