Depression and Anxiety Prediction Using Deep Language Models and Transfer Learning

Feasibility of a Machine Learning-Based Smartphone Application in Detecting Depression and Anxiety in a Generally Senior Population

“…Usage statistics and survey results from this study taken together indicate that utilizing regular voice recordings of users answering questions from a smartphone app to analyze their levels of anxiety and depression is feasible. Ellipsis Health has previously published results of semantic ( Rutowski et al, 2019 , 2020 ) and acoustic ( Harati et al, 2021 ) analysis of speech to detect depression and anxiety using models trained, to the best of our knowledge, with the largest database reported in the literature ( Rutowski et al, 2019 ). We have also previously reported this algorithm performance is maintained (i.e., is portable) when applied to the current study population using long short-term memory (LSTM) models ( Rutowski et al, 2020 ).…”

“…Ellipsis Health has previously published results of semantic ( Rutowski et al, 2019 , 2020 ) and acoustic ( Harati et al, 2021 ) analysis of speech to detect depression and anxiety using models trained, to the best of our knowledge, with the largest database reported in the literature ( Rutowski et al, 2019 ). We have also previously reported this algorithm performance is maintained (i.e., is portable) when applied to the current study population using long short-term memory (LSTM) models ( Rutowski et al, 2020 ). The newer transformer methodology was also portable (performance was within 10% when comparing AUCs of the original training dataset and this study’s dataset).…”

“…In ongoing work, large corpora of transcribed speech are used for natural language processing (NLP) training to further develop semantic speech analysis ( Lan et al, 2019 ; Yang et al, 2019 ). The most relevant NLP advancements have been used for transfer learning and improvements in deep learning architecture like transformers ( Bengio, 2012 ; Vaswani et al, 2017 ), which maintain performance without using prohibitive amounts of labeled data ( Rutowski et al, 2020 ; Harati et al, 2021 ).…”

Evaluating the Feasibility and Acceptability of an Artificial-Intelligence-Enabled and Speech-Based Distress Screening Mobile App for Adolescents and Young Adults Diagnosed with Cancer: A Study Protocol

“…Second, a recent systematic review found that no tool so far was equipped to evaluate D&A simultaneously. About half of the studies on speech-based mental health assessment focused on depression, but only a small number (5%) examined anxiety, whereas studies of both conditions are rare [ 36 , 38 ]. It is problematic because depression and anxiety often co-occur, and anxiety disorders are highly prevalent among AYA survivors, e.g., health anxiety or fear of reoccurrence.…”

“…The models that underlie the EH Voice Tool show top performance in comparison to results published in the literature [ 44 ]. Specifically, results for the EH Voice Tool’s performance on PHQ/GAD prediction are currently 0.85/0.84 for area under the curve (AUC; for binary classification with a threshold of 10), 4.25/4.47 for root mean square error (RMSE, a measure of average regression error), and 3.13/3.23 for mean absolute error (MAE, another measure of average regression error) [ 38 , 40 ]. Finally, to the best of our knowledge, the EH Voice Tool is the only speech-based distress screening and monitoring tool that simultaneously generates results for depression and anxiety scores, an essential strategy for reducing fatigue over long-term and repetitive screening.…”