AUTHOR=Li Nanxi , Feng Lei , Hu Jiaxue , Jiang Lei , Wang Jing , Han Jiali , Gan Lu , He Zhiyang , Wang Gang 

TITLE=Using deeply time-series semantics to assess depressive symptoms based on clinical interview speech

JOURNAL=Frontiers in Psychiatry

VOLUME=Volume 14 - 2023

YEAR=2023

URL=https://www.frontiersin.org/journals/psychiatry/articles/10.3389/fpsyt.2023.1104190

DOI=10.3389/fpsyt.2023.1104190

ISSN=1664-0640

ABSTRACT=Depression is an affect disorder that contributes to a significant global disease burden. Measurement-Based Care (MBC) during full course management is advocated, with symptom assessment being one of the important components. Current depressive symptoms evaluation primarily depends on face-to-face clinical interviews. Rating scales are widely used as a convenient and powerful tool; however, they are influenced by the subjectivity and consistency of the raters. The assessment of depressive symptoms is usually carried out with a clear purpose and restricted content, such as clinical interviews based on Hamilton Depression Rating Scale (HAMD); thus, the corresponding results are easily obtained and quantified. Artificial intelligence (AI) techniques are used due to their objective, stable, and consistent performance and are suitable for assessing depressive symptoms. Therefore, this study applied Deep Learning (DL)-based Natural Language Processing (NLP) techniques to assess depressive symptoms during clinical interviews; thus, we proposed an algorithm model, explored the techniques feasibility, and evaluated its performance. This study enrolled 329 patients with Major Depressive Episode. Clinical interviews based on the HAMD-17 were conducted by trained psychiatrists, with their speech being recorded simultaneously. A total of 387 audio recordings were included in the final analysis. A deeply time-series semantics model for the assessment of depressive symptoms based on multi-granularity and multi-task joint training (MGMT) is proposed. The results show that the MGMTs performance is acceptable for assessing depressive symptoms. Furthermore, the MGMT has an F1 score (a metric to measure model performance) of 0.719 in the classification of the four-level severity of depression and 0.890 in identifying the presence of depressive symptoms. This also demonstrates the feasibility of the DL and the NLP techniques applied to the clinical interview and the assessment of depressive symptoms. However, there are limitations to this study including the lack of an adequate sample size during the early stage of research, as well as using speech content alone to assess depressive symptoms will lose the information gained through observation. A multi-dimensional model combing semantics with speech voice, facial expression, and other valuable information, as well as taking into account personalized information is a potential direction in the future.