Facial expression recognition to predict antidepressant medication outcome
Depression is a leading cause of disability: over 350 million people in the world today have a reduced ability to work and function socially. Nearly 50 % of sufferers fail to respond to the first antidepressant prescribed, prolonging the period of negativity on life events.
A web-based tool for recognising emotions
The EU-funded PReDicT project has implemented innovative digital technologies to better diagnose mental health problems. i-spero®(opens in new window) is a web-based digital tool that includes a number of functionalities including diagnostic assessment and triage, ongoing monitoring of symptoms, and provision of real-time health alerts to both patients and physicians. The i-spero® platform includes the PReDicT(opens in new window) test which harnesses the power of machine learning to quantify a patient’s negative outlook and perceptions. It works by asking patients to recognise a facial expression they see on their smart phone, tablet, or PC for example happy, sad or fearful. As depressed patients are better able to identify expressions associated with negative emotions, the recognition of subtle facial expressions before and after short-term treatment can predict if the patient’s condition will improve or if a change in their medical intervention is required. “Our goal was to get patients on to the right treatment sooner than with existing methods,” explains project coordinator Rebecca Dias. To evaluate the effectiveness of PReDicT, project partners conducted a large-scale international clinical study recruiting over 900 patients with depression from 5 major European countries. The investigation systematically measured if the antidepressant medication changed how patients saw facial expressions. Patients that did not exhibit any improvement were switched to a higher dose of the antidepressant drug or to a different drug within 7 days of starting treatment. The control group was treated in the conventional way, with their physician waiting 4-6 weeks after treatment to assess medication efficacy. In addition, PReDicT measured compliance rates, both in terms of patients’ adherence to their treatment and study plan, but also with respect to using digital assessments and the test as part of their routine care. PReDicT improved key patient outcomes, specifically by reducing levels of anxiety and accelerating functional recovery by 6 months, compared to the control group. “Our approach of gathering information to support proactive rather than reactive clinical decision making received overwhelmingly positive feedback from both patients and physicians,” emphasises Dias.
Project significance and future prospects
PReDicT was an international collaboration that successfully integrated digital technology into primary care across Europe. At the same time, the project has highlighted differences in treatment, prescription patterns and outcome measures of depression across different European countries. Currently there is no similar tool that matches the specificity and sensitivity of PReDicT. Preliminary results already show societal cost savings and, alongside the improvement in patient wellbeing, they are expected to facilitate the fast integration of PReDicT into routine clinical practice. Partners are already testing the device in United Kingdom healthcare centres and are also looking to incorporate i-spero® as a companion diagnostic to pharmaceutical companies developing new therapies for the treatment of depression.
