PS54: AI-Driven Medical Imaging: Promoting Innovation as Drug and Medical Device Development Tools

The rapid evolution of the digital era, characterized by large scale AI models, multimodal data integration, and automated analytical pipelines, is reshaping how clinical development decisions are made. Once primarily qualitative and labor intensive, medical imaging has become a rich and quantifiable digital data source that increasingly supports the development and qualification of drug and medical device development tools to influence key decision-making, from disease diagnosis, endpoint selection, early signal detection to regulatory submission. The year of 2025 marked two pivotal FDA regulatory milestones in medical imaging: the De Novo authorization to Clairity Breast, the first AI-powered prognostic platform to predict breast cancer risk, and the qualification of the first imaging based drug development tool (DDT) in metabolic dysfunction-associated steatohepatitis (MASH) clinical trials. These landmark decisions reflect the Agency's growing confidence in imaging derived and quantitatively validated evidence as decision-grade development tools across drugs and devices. They also underscore the urgent need for transdisciplinary collaboration among statisticians, imaging scientists, AI/ML developers, clinicians, and regulators on building transparent, reproducible, and interpretable imaging pipelines that are fit for context of use and aligned with FDA qualification frameworks. At this particularly timely juncture, this session will explore how AI enabled medical imaging, combined with modern statistical science, is emerging as a critical transdisciplinary tool for impactful decision making in clinical trials through its role as a fit-for-purpose drug and medical device development tool. Clinical studies now routinely generate multi-modal imaging data that are integrated with clinical, biomarker, and real world data, supporting imaging-derived endpoint quantification within well-defined contexts of use. This creates ample opportunities for optimizing impactful decision-making, as well as methodological and regulatory challenges around reliability, interpretability, and risk evaluation that are central to regulatory qualification and acceptance. Bringing together several experts from the FDA, the pharmaceutical industry, and academia, this session highlights recent advances and current thinking at the intersection of AI and statistical science, with a particular emphasis on AI-enabled medical imaging for clinical development as an integrated ecosystem of drug and medical device development tools. The presentations involve various transdisciplinary decision-making (i.e., pathology-based endpoint quantification, patient response assessment and synthetic data generation) from diverse data sources (i.e., clinical trials, registries and electronic medical records [EMR]) and illustrate how cross-disciplinary collaboration accelerates tool development, validation, and regulatory acceptance. Collectively, this session aims to address a fast moving regulatory and technological frontier, foster cross sector dialogue, inspire methodological innovation, and demonstrate how AI and statistical science play a central role in ensuring AI driven imaging tools to provide decision-grade evidence and promote innovation through the successful development and qualification of drug and medical device development tools in today's rapidly evolving digital environment.