Detection of major depressive disorder using vocal acoustic analysis and machine learning—an exploratory study

Abstract

Diagnosis and treatment in psychiatry are still highly dependent on reports from patients and on clinician judgment. This fact makes them prone to memory and subjectivity biases. As for other medical fields, where objective biomarkers are available, there has been an increasing interest in the development of such tools in psychiatry. To this end, vocal acoustic parameters have been recently studied as possible objective biomarkers, instead of otherwise invasive and costly methods. Patients suffering from different mental disorders, such as major depressive disorder (MDD), may present with alterations of speech. These can be described as uninteresting, monotonous, and spiritless speech and low voice. Thirty-three individuals (11 males) over 18 years old were selected, 22 of which being previously diagnosed with MDD and 11 healthy controls. Their speech was recorded in naturalistic settings, during a routine medical evaluation for psychiatric patients, and in different environments for healthy controls. Voices from third parties were removed. The recordings were submitted to a vocal feature extraction algorithm, and to different machine learning classification techniques. The results showed that random tree models with 100 trees provided the greatest classification performances. It achieved mean accuracy of 87.5575% ± 1.9490, mean kappa index, sensitivity, and specificity of 0.7508 ± 0.0319, 0.9149 ± 0.0204, and 0.8354 ± 0.0254, respectively, for the detection of MDD. The use of machine learning classifiers with vocal acoustic features appears to be very promising for the detection of major depressive disorder in this exploratory study, but further experiments with a larger sample will be necessary to validate our findings.