SBIR Phase I: Scaling Mental Healthcare in COVID-19 with Voice Biomarkers

小型企业创新研究(SBIR)第一阶段项目的更广泛影响/商业潜力是使用语音作为精神健康的实时测量。将语音语调转换为生物标志物可以实现疾病诊断和病情发展，支持价值130亿美元的虚拟医疗保健部门，在新冠肺炎出现之前，该部门的年增长率为27%。此外，同伴对精神健康的支持增加了自我护理的参与度，减少了物质使用和抑郁，特别是对脆弱人群来说。该项目将在小组环境中推进机器学习在语音心理健康生物标记物中的使用。这个小型企业创新研究(SBIR)第一阶段项目将为基于深度强化学习的系统定义语音生物标记物特征。该项目将推进语音生物标记技术，作为快速行为健康诊断，有可能取代目前纸质的PHQ-9和GAD-7测试。当务之急是根据最大限度地改善抑郁和焦虑评分的奖励功能，衡量团体治疗的个人和活动的最佳组合。主要的技术挑战包括：(1)捕获视频中的非语言线索；(2)解释多人音频处理；(3)创建深度强化学习模型以服务于相关的小组比赛和后续练习；以及(4)从小组会议中建立引人入胜的进度视觉反馈。这项创新的预期技术成果将是在临床相关环境中为基于深度强化学习的系统定义语音生物标记物特征和奖励功能，以改善抑郁症和焦虑的治疗结果。该奖项反映了NSF的法定使命，并已通过使用基金会的智力优点和更广泛的影响审查标准进行评估，被认为值得支持。

The broader impact/commercial potential of this Small Business Innovation Research (SBIR) Phase I project is to use voice as a real-time measurement of mental health. Transforming voice intonations into biomarkers could enable disease diagnosis and progression, supporting the $13 B virtual health care sector that was growing 27% annually prior to COVID-19. Furthermore, peer support for mental health increases engagement in self-care decreases substance use and depression, particularly for vulnerable populations. The project will advance the use of machine learning for voice mental health biomarkers in a group setting. This Small Business Innovation Research (SBIR) Phase I project will define voice biomarker features for a deep reinforcement learning based system. This project will advance a voice biomarker technology that can serve as fast behavioral health diagnostic, potentially superseding the current paper-based PHQ-9 and GAD-7 tests. The priority is to scale the optimal mix of individuals and activities for group therapy based on reward functions that maximize improvements in depression and anxiety scores. The major technical challenges include: (1) capturing nonverbal cues in a video; (2) interpreting multi-speaker audio processing; (3) creating deep reinforcement learning models to serve relevant group matches and follow-up exercises; and (4) building engaging visual feedback of progress from group meetings. The anticipated technical result of this innovation will be to define voice biomarker features and reward functions for a deep reinforcement learning based system in clinically relevant settings to improve depression and anxiety treatment outcomes.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.