Differentiation between bipolar versus unipolar depression by AI-based multimodal behavioral analysis

Given their differing pathophysiology and treatment approaches, the primary aim of the current study was to develop a methodology accurately distinguishing bipolar from unipolar depression.

Adults with bipolar disorder or major depressive disorder were recruited nationwide in Japan and assessed via recorded interviews. During the interview, diagnoses were established using the Structured Clinical Interview for DSM-5-Research Version. Depressive and manic symptom severity was assessed using the Hamilton Depression Rating Scale (HDRS) and Young Mania Rating Scale (YMRS), respectively. Artificial intelligence (AI)-based integrative, quantitative analyses of behavioral characteristics, including facial expressions, eye gaze, and voice prosody, were conducted on the recorded video.

Between August 2022 and March 2024, 379 participants completed interviews, of whom 366 were analyzed (120 bipolar I, 93 bipolar II, 122 major depressive, and 74 persistent depressive disorder). AI-based diagnostic differentiation between bipolar and unipolar depression using four-fold, three-times repeated cross-validation achieved an overall accuracy of 78%, with an area under the curve of 0.76. Key features contributing to bipolar diagnosis included greater temporal instabilities in neutral facial expression, eye gaze, sad facial expression, and root mean square amplitude of speech. HDRS could be predicted in unipolar and bipolar depression, with influential predictors of depression including sleep inefficiency, autistic traits, variability in sad facial expression, and comorbid anxiety disorder. YMRS could be predicted with major predictors including comorbid anxiety disorder, sleep inefficiency, and variabilities in sad facial expression and eye movement.

Integrating multimodal behavioral indicators with AI-based modeling enables clinically meaningful differentiation between bipolar and unipolar depression.