Mental Illness Detection and Clinical Assessment with Reliable Interpretability

ongoing

"Mental health represents a major worry in most countries while providing a good diagnosis is a hard task. A reliable system for assessing mental disorders is crucial for helping clinicians and patients achieve accurate diagnoses and treatment. Research has long established that certain vocal features are linked to mental health disorders, allowing a machine learning model to learn how to classify different mental disorders using voice. Artificial intelligence (AI) has been providing excellent responses diagnosing mental diseases so far but, remarkably, they often operate as ""black boxes"" that do not allow for an understanding of how they make their decisions. This makes it mandatory to perform a thorough study on the system in order to achieve fairness, explainability and trustworthiness. This project aims to develop an open source AI based model, explainable and transparent, that can assess between many mental illnesses using voice recordings from a patient, contributing from the study, analysis and understanding of the problem, providing greater interpretability, security and extrapolation ability. The project will be divided in three stages, curation of dataset, the development of the system and exhaustive explainability and fairness analysis over the system. This tool does not replace a clinician diagnosis, it intends to help them decide in dubious cases, with non invasive, cheap and easy tests that can be performed with a cell phone by the patient or his/her family to keep a daily control. This is a multidisciplinary project that combines health sciences, machine learning and mathematical modelling. This project also intends to share expertise with the host organization and equip the candidate with training advanced methodologies. The results could substantially enhance the effectiveness of mental health diagnosing. This project aligns with the EU’s strategy for safeguarding public health."