FDA Accepts First AI-Based Neuroscience Drug-Development Tool

The FDA’s Center for Drug Evaluation and Research (CDER) accepted a submission into the Innovative Science and Technology Approaches for New Drugs (ISTAND) pilot that represents the first AI-based and digital health technology project in neuroscience. The submission is from Deliberate AI and concerns an AI-generated Clinical Outcome Assessment (AI-COA) designed to quantify anxiety and depression severity from multimodal behavioral and physiologic signals.

ISTAND, created in 2020 to provide a review pathway for novel drug-development tools that fall outside existing evaluation routes, began accepting submissions in 2022. The ISTAND pilot is explicitly intended to facilitate qualification of tools that can improve drug development efficiency and endpoint measurement. Deliberate AI’s AI-COA applies multimodal behavioral signal processing and machine learning on inputs including facial expressions, speech behaviors and acoustics, physical movements, pupil changes and vital signs to build quantitative assessments of mental-health symptoms.

CDER accepted Deliberate AI’s letter of intent, confirming the tool is feasible, practical and addresses a specific need in drug development. Per FierceBiotech coverage, the letter of intent acceptance advances the project to the next of two qualification steps; Deliberate AI will work with the FDA to develop a formal qualification plan. Peter Stein, M.D., director of the Office of New Drugs at CDER, called the move “a pioneering step for the ISTAND program as the first artificial intelligence-based, digital health technology project in neuroscience to be accepted into the pilot program,” and linked the acceptance to FDA’s goal of optimizing drug development and potentially expediting access to safe, effective treatments.

This is an early regulatory precedent for AI/ML-driven digital clinical outcome assessments and digital biomarkers in psychiatric indications. Acceptance into ISTAND means the FDA is willing to engage on qualification pathways for complex, multimodal ML tools as drug-development assets. For ML engineers, data scientists, and regulatory teams building similar tools, the next milestone — the qualification plan and its validation/data requirements — will be a critical signal about acceptable evidence standards, data modalities, labeling approaches and endpoints.

the content of the qualification plan, the evidence-generation and validation strategy the FDA requires, and any published guidance or qualification documentation in the CDER & CBER Drug Development Tool Qualification Project Search. These artifacts will set practical expectations for data collection, model validation, and deployment constraints for AI-driven digital COAs.