It’s not easy to get depression-detecting AI through the FDA

Kintsugi, a California-based startup, spent seven years developing AI capable of detecting signs of depression and anxiety through speech analysis. However, the company is ceasing operations after failing to obtain FDA clearance for its technology. Kintsugi’s software analyzes speech patterns – pauses, sentence structure, and speed – to identify subtle indicators of mental health issues, aiming to complement traditional screening methods like the PHQ-9 questionnaire. The company sought approval through the FDA’s De Novo pathway, but faced challenges due to the agency’s unfamiliarity with AI and a regulatory framework designed for static devices, hindering the ability to continuously update the AI model. Despite efforts, funding ran out during the lengthy approval process. Kintsugi is now open-sourcing most of its technology, hoping others will continue the work, while retaining some technology for detecting deepfake audio. Concerns exist regarding potential misuse of the open-sourced mental health screening model, but the founder believes underuse is a more realistic risk. The company’s experience highlights the difficulties startups face navigating medical regulation and the need for a more streamlined approval process for AI-based medical devices.

(Source：The Verge)