Automatic Assessment of Emotion Dysregulation in American, French, and Tunisian Adults and New Developments in Deep Multimodal Fusion: Cross-sectional Study

Abstract

Background Emotion dysregulation is a key dimension of adult psychological functioning. There is an interest in developing a computer-based, multimodal, and automatic measure. Objective We wanted to train a deep multimodal fusion model to estimate emotion dysregulation in adults based on their responses to the Multimodal Developmental Profile, a computer-based psychometric test, using only a small training sample and without transfer learning. Methods Two hundred and forty-eight participants from 3 different countries took the Multimodal Developmental Profile test, which exposed them to 14 picture and music stimuli and asked them to express their feelings about them, while the software extracted the following features from the video and audio signals: facial expressions, linguistic and paralinguistic characteristics of speech, head movements, gaze direction, and heart rate variability derivatives. Participants also responded to the brief version of the Difficulties in Emotional Regulation Scale. We separated and averaged the feature signals that corresponded to the responses to each stimulus, building a structured data set. We transformed each person’s per-stimulus structured data into a multimodal codex, a grayscale image created by projecting each feature’s normalized intensity value onto a cartesian space, deriving each pixel’s position by applying the Uniform Manifold Approximation and Projection method. The codex sequence was then fed to 2 network types. First, 13 convolutional neural networks dealt with the spatial aspect of the problem, estimating emotion dysregulation by analyzing each of the codified responses. These convolutional estimations were then fed to a transformer network that decoded the temporal aspect of the problem, estimating emotional dysregulation based on the succession of responses. We introduce a Feature Map Average Pooling layer, which computes the mean of the convolved feature maps produced by our convolution layers, dramatically reducing the number of learnable weights and increasing regularization through an ensembling effect. We implemented 8-fold cross-validation to provide a good enough estimation of the generalization ability to unseen samples. Most of the experiments mentioned in this paper are easily replicable using the associated Google Colab system. Results We found an average Pearson correlation (r) of 0.55 (with an average P value of <.001) between ground truth emotion dysregulation and our system’s estimation of emotion dysregulation. An average mean absolute error of 0.16 and a mean concordance correlation coefficient of 0.54 were also found. Conclusions In psychometry, our results represent excellent evidence of convergence validity, suggesting that the Multimodal Developmental Profile could be used in conjunction with this methodology to provide a valid measure of emotion dysregulation in adults. Future studies should replicate our findings using a hold-out test sample. Our methodology could be implemented more generally to train deep neural networks where only small training samples are available.

Modern views of machine learning for precision psychiatry

“…302 The success of transformer networks in jointly modeling video, speech, and language data has catalyzed multi-modal modeling in mental health. 303 Multi-modal modeling techniques can also be used in characterizing symptoms such as emotion dysregulation, 304 loneliness, 305 and sentiment analysis. 306 In a prognostic study, an SVM-based multi-modal ML approach was developed to integrate clinical, neurocognitive, neuroimaging, and genetic information to predict psychosis in patients with clinical high-risk states.…”

Modern Views of Machine Learning for Precision Psychiatry

“…The success of transformer networks in jointly modeling video, speech, and language data has catalyzed multimodal modeling in mental health [295]. Multimodal modeling techniques can also be used in modeling symptoms such as emotion dysregulation [296], loneliness [297] and sentiment analysis [298]. In a prognostic study, an SVM-based multimodal ML approach was developed to integrate clinical, neurocognitive, neuroimaging, and genetic information to predict psychosis in patients with clinical high-risk states [76].…”

Modern Views of Machine Learning for Precision Psychiatry

“…The success of transformer networks in jointly modeling video, speech, and language data has catalyzed multimodal modeling in mental health [296]. Multimodal modeling techniques can also be used in modeling symptoms such as emotion dysregulation [297], loneliness [298] and sentiment analysis [299]. In a prognostic study, an SVM-based multimodal ML approach was developed to integrate clinical, neurocognitive, neuroimaging, and genetic information to predict psychosis in patients with clinical high-risk states [76].…”