Design and Development of Depression Detection studies using Multimodal Digital Biomarkers and Machine Learning

Early detection of depression is important for the prevention of its adverse impact on human lives and to ensure timely and appropriate medical intervention. For depression detection, current clinical practices include standardized interviews and self-reporting questionnaires. However, there are major challenges associated, such as self-reporting leading to bias, inability to differentiate between different mental disorders, challenges faced by children or people with neurological disorders and, high dependence on clinical expertise to validate the self-reporting. This reduces the possibility of early intervention and detection. One promising method for depression screening/diagnosis is to use machine learning (ML) on the passive data collected using the wearable devices. Various digital biomarkers can be employed for developing the ML based techniques and predicting the depression onset. However, unavailability of the required data, lack of standardized data collection protocols and lack of multidisciplinary approach are certain challenges. In this work, our aim is to combine the complimentary expertise of the networking partners in exploring the use of digital biomarkers and passive data streams to predict and screen for depression. With the help of available primary data with the networking partners, we intend to identify the appropriate multimodal biomarkers and aim to develop a suitable experimental design protocol that will form the standardized mechanism for primary data collection as a part of a future research grant. The ACME networking grant will serve as a seed funding for developing the experimental protocols and explore the commercial development plan that can eventually be converted into a viable digital product.