Privacy-preserving and scalable voice-based tools for cognitive assessment - PROJECT SUMMARY This project aims to enhance and disseminate artificial intelligence (AI) tools for analysis of digital voice recorded assessments, and utilizing the Global Research and Imaging Platform (GRIP) to ensure scalability, security, sustainability, and broad access. Early detection of cognitive impairment is crucial for timely interventions. Current diagnostic tools, such as neuropsychological tests and self-reported questionnaires, are often biased and imprecise. Voice recordings, however, contain rich indices that, when analyzed using AI models, can detect early signs of cognitive decline. Our team has developed and validated AI models using large datasets of voice recordings collected during neuropsychological assessments. These models have shown strong potential for detecting early cognitive impairment. However, barriers such as the tools’ scalability, interoperability, and privacy concerns related to the sharing of personally identifiable information (PII) in voice data have limited broader dissemination and adoption. This project will address these challenges by refactoring the AI tools using best practices in software development to ensure robustness, security, portability, and scalability. Specifically, we will develop modular, open-source software that can operate efficiently in cloud environments, enabling researchers and clinicians to integrate the tools into their own workflows. Additionally, we will create privacy-preserving pipelines that automatically obfuscate voice data while retaining the necessary features for cognitive assessment, ensuring secure sharing of data in compliance with regulatory standards. These pipelines and tools will be integrated into GRIP’s platform, a modular, open-access infrastructure that supports best practices in software dissemination. The final output will include a fully documented suite of tools for analyzing voice data, accessible to researchers with varying levels of technical expertise. By making these tools freely available through GRIP, we aim to foster broad community engagement, including training workshops and webinars to encourage adoption. This project aligns with NIH’s goals of enhancing the sustainability and impact of research software by promoting robust, reusable, and scalable tools that adhere to Findable, Accessible, Interoperable, and Reusable for Principles for Research Software (FAIR4RS) principles. The dissemination of these tools will support early diagnosis of cognitive decline and monitoring of cognitive status, ultimately improving patient outcomes.
