Multimodal Detection of Depression in Clinical Interviews

Abstract

Current methods for depression assessment depend almost entirely on clinical interview or selfreport ratings. Such measures lack systematic and efficient ways of incorporating behavioral observations that are strong indicators of psychological disorder. We compared a clinical interview of depression severity with automatic measurement in 48 participants undergoing treatment for depression. Interviews were obtained at 7-week intervals on up to four occasions. Following standard cut-offs, participants at each session were classified as remitted, intermediate, or depressed. Logistic regression classifiers using leave-one-out validation were compared for facial movement dynamics, head movement dynamics, and vocal prosody individually and in combination. Accuracy (remitted versus depressed) for facial movement dynamics was higher than that for head movement dynamics; and each was substantially higher than that for vocal prosody. Accuracy for all three modalities together reached 88.93%, exceeding that for any single modality or pair of modalities. These findings suggest that automatic detection of depression from behavioral indicators is feasible and that multimodal measures afford most powerful detection.

Dynamic Multimodal Measurement of Depression Severity Using Deep Autoencoding

“…To further assess the quality of the proposed dynamic feature sets, we compared them with the dynamic features proposed in our earlier work [41] and with alternative dynamic features that include facial movement [14], head movement [19], or prosodic features [11]. For a fair and accurate comparison between the proposed dynamic features and alternative methods, it was necessary to re-implement the alternative methods for: (1) the more challenging problem of measurement of 3-levels of depression severity (as compared to 2-classes classification), and (2) evaluating them on our clinical data.…”

“…For a fair and accurate comparison between the proposed dynamic features and alternative methods, it was necessary to re-implement the alternative methods for: (1) the more challenging problem of measurement of 3-levels of depression severity (as compared to 2-classes classification), and (2) evaluating them on our clinical data. Thus, we re-implemented previous methods as well as could be done from their description in the corresponding papers (including adapting our own work [41] to the problem of 3-levels of depression severity).…”

“…4.d). To this end, we used two different representations to describe the videos with a fixed length representation: 1) Improved Fisher Vector (IFV) coding [54], and 2) Compact Dynamic Feature Set (DFS) [41]. …”

“…Part of this work has been presented at the ACM International Conference on Multimodal Interaction [41]. …”

Automatic Measurement of Head and Facial Movement for Analysis and Detection of Infants’ Positive and Negative Affect

“…Girard and colleagues , in a longitudinal study of adult patients in treatment for depression, found that head motion velocity increased as they recovered from depression. Dibeklioglu and colleagues (Dibeklioglu et al, 2015) reported related findings for both head and face. In addition to communicating affect, head motion serves regulatory functions.…”