Identifying depression-related topics in smartphone-collected free-response speech recordings using an automatic speech recognition system and a deep learning topic model.

BACKGROUND: Prior research has associated spoken language use with depression, yet studies often involve small or non-clinical samples and face challenges in the manual transcription of speech. This paper aimed to automatically identify depression-related topics in speech recordings collected from clinical samples. METHODS: The data included 3919 English free-response speech recordings collected via smartphones from 265 participants with a depression history. We transcribed speech recordings via automatic speech recognition (Whisper tool, OpenAI) and identified principal topics from transcriptions using a deep learning topic model (BERTopic). To identify depression risk topics and understand the context, we compared participants' depression severity and behavioral (extracted from wearable devices) and linguistic (extracted from transcribed texts) characteristics across identified topics. RESULTS: From the 29 topics identified, we identified 6 risk topics for depression: 'No Expectations', 'Sleep', 'Mental Therapy', 'Haircut', 'Studying', and 'Coursework'. Participants mentioning depression risk topics exhibited higher sleep variability, later sleep onset, and fewer daily steps and used fewer words, more negative language, and fewer leisure-related words in their speech recordings. LIMITATIONS: Our findings were derived from a depressed cohort with a specific speech task, potentially limiting the generalizability to non-clinical populations or other speech tasks. Additionally, some topics had small sample sizes, necessitating further validation in larger datasets. CONCLUSION: This study demonstrates that specific speech topics can indicate depression severity. The employed data-driven workflow provides a practical approach for analyzing large-scale speech data collected from real-world settings.

HEAR4Health: a blueprint for making computer audition a staple of modern healthcare.

Recent years have seen a rapid increase in digital medicine research in an attempt to transform traditional healthcare systems to their modern, intelligent, and versatile equivalents that are adequately equipped to tackle contemporary challenges. This has led to a wave of applications that utilise AI technologies; first and foremost in the fields of medical imaging, but also in the use of wearables and other intelligent sensors. In comparison, computer audition can be seen to be lagging behind, at least in terms of commercial interest. Yet, audition has long been a staple assistant for medical practitioners, with the stethoscope being the quintessential sign of doctors around the world. Transforming this traditional technology with the use of AI entails a set of unique challenges. We categorise the advances needed in four key pillars: Hear, corresponding to the cornerstone technologies needed to analyse auditory signals in real-life conditions; Earlier, for the advances needed in computational and data efficiency; Attentively, for accounting to individual differences and handling the longitudinal nature of medical data; and, finally, Responsibly, for ensuring compliance to the ethical standards accorded to the field of medicine. Thus, we provide an overview and perspective of HEAR4Health: the sketch of a modern, ubiquitous sensing system that can bring computer audition on par with other AI technologies in the strive for improved healthcare systems.

Multilingual markers of depression in remotely collected speech samples: A preliminary analysis.

BACKGROUND: Speech contains neuromuscular, physiological and cognitive components, and so is a potential biomarker of mental disorders. Previous studies indicate that speaking rate and pausing are associated with major depressive disorder (MDD). However, results are inconclusive as many studies are small and underpowered and do not include clinical samples. These studies have also been unilingual and use speech collected in controlled settings. If speech markers are to help understand the onset and progress of MDD, we need to uncover markers that are robust to language and establish the strength of associations in real-world data. METHODS: We collected speech data in 585 participants with a history of MDD in the United Kingdom, Spain, and Netherlands as part of the RADAR-MDD study. Participants recorded their speech via smartphones every two weeks for 18 months. Linear mixed models were used to estimate the strength of specific markers of depression from a set of 28 speech features. RESULTS: Increased depressive symptoms were associated with speech rate, articulation rate and intensity of speech elicited from a scripted task. These features had consistently stronger effect sizes than pauses. LIMITATIONS: Our findings are derived at the cohort level so may have limited impact on identifying intra-individual speech changes associated with changes in symptom severity. The analysis of features averaged over the entire recording may have underestimated the importance of some features. CONCLUSIONS: Participants with more severe depressive symptoms spoke more slowly and quietly. Our findings are from a real-world, multilingual, clinical dataset so represent a step-change in the usefulness of speech as a digital phenotype of MDD.

Long-term participant retention and engagement patterns in an app and wearable-based multinational remote digital depression study.

Recent growth in digital technologies has enabled the recruitment and monitoring of large and diverse populations in remote health studies. However, the generalizability of inference drawn from remotely collected health data could be severely impacted by uneven participant engagement and attrition over the course of the study. We report findings on long-term participant retention and engagement patterns in a large multinational observational digital study for depression containing active (surveys) and passive sensor data collected via Android smartphones, and Fitbit devices from 614 participants for up to 2 years. Majority of participants (67.6%) continued to remain engaged in the study after 43 weeks. Unsupervised clustering of participants' study apps and Fitbit usage data showed 3 distinct engagement subgroups for each data stream. We found: (i) the least engaged group had the highest depression severity (4 PHQ8 points higher) across all data streams; (ii) the least engaged group (completed 4 bi-weekly surveys) took significantly longer to respond to survey notifications (3.8 h more) and were 5 years younger compared to the most engaged group (completed 20 bi-weekly surveys); and (iii) a considerable proportion (44.6%) of the participants who stopped completing surveys after 8 weeks continued to share passive Fitbit data for significantly longer (average 42 weeks). Additionally, multivariate survival models showed participants' age, ownership and brand of smartphones, and recruitment sites to be associated with retention in the study. Together these findings could inform the design of future digital health studies to enable equitable and balanced data collection from diverse populations.

Biopsychosocial Response to the COVID-19 Lockdown in People with Major Depressive Disorder and Multiple Sclerosis.

BACKGROUND: Changes in lifestyle, finances and work status during COVID-19 lockdowns may have led to biopsychosocial changes in people with pre-existing vulnerabilities such as Major Depressive Disorders (MDDs) and Multiple Sclerosis (MS). METHODS: Data were collected as a part of the RADAR-CNS (Remote Assessment of Disease and Relapse-Central Nervous System) program. We analyzed the following data from long-term participants in a decentralized multinational study: symptoms of depression, heart rate (HR) during the day and night; social activity; sedentary state, steps and physical activity of varying intensity. Linear mixed-effects regression analyses with repeated measures were fitted to assess the changes among three time periods (pre, during and post-lockdown) across the groups, adjusting for depression severity before the pandemic and gender. RESULTS: Participants with MDDs (N = 255) and MS (N = 214) were included in the analyses. Overall, depressive symptoms remained stable across the three periods in both groups. A lower mean HR and HR variation were observed between pre and during lockdown during the day for MDDs and during the night for MS. HR variation during rest periods also decreased between pre- and post-lockdown in both clinical conditions. We observed a reduction in physical activity for MDDs and MS upon the introduction of lockdowns. The group with MDDs exhibited a net increase in social interaction via social network apps over the three periods. CONCLUSIONS: Behavioral responses to the lockdown measured by social activity, physical activity and HR may reflect changes in stress in people with MDDs and MS. Remote technology monitoring might promptly activate an early warning of physical and social alterations in these stressful situations. Future studies must explore how stress does or does not impact depression severity.

Fitbeat: COVID-19 estimation based on wristband heart rate using a contrastive convolutional auto-encoder.

This study proposes a contrastive convolutional auto-encoder (contrastive CAE), a combined architecture of an auto-encoder and contrastive loss, to identify individuals with suspected COVID-19 infection using heart-rate data from participants with multiple sclerosis (MS) in the ongoing RADAR-CNS mHealth research project. Heart-rate data was remotely collected using a Fitbit wristband. COVID-19 infection was either confirmed through a positive swab test, or inferred through a self-reported set of recognised symptoms of the virus. The contrastive CAE outperforms a conventional convolutional neural network (CNN), a long short-term memory (LSTM) model, and a convolutional auto-encoder without contrastive loss (CAE). On a test set of 19 participants with MS with reported symptoms of COVID-19, each one paired with a participant with MS with no COVID-19 symptoms, the contrastive CAE achieves an unweighted average recall of 95.3 % , a sensitivity of 100 % and a specificity of 90.6 % , an area under the receiver operating characteristic curve (AUC-ROC) of 0.944, indicating a maximum successful detection of symptoms in the given heart rate measurement period, whilst at the same time keeping a low false alarm rate.

An arterial wall motion test phantom for the evaluation of wall motion software.

Increasing cardiovascular disease has led to new ultrasound methods of assessing artery disease such as arterial wall motion measurement. To validate arterial wall motion software, we developed a mechanically-controlled wall motion test phantom with straight upper and lower agar tissue mimicking material layers that resemble an artery cross section. The wall separation, displacements and wall velocities and accelerations can be controlled within physiologically realistic levels. A user-definable displacement or one of several pre-defined displacement waveforms can be created by the user with custom-written software. The test phantom is then controlled using the defined waveform with a stepper motor controller. Accuracy assessment of the test phantom with a laser vibrometer yielded a positional accuracy of 36+/-2 microm. A typical application of the test phantom is demonstrated by assessing a tissue Doppler imaging (TDI) method for estimating the distension waveform. The TDI-based method was found to have a minimum resolvable displacement of 22.5 microm, and a measurement accuracy of +/-8% using a physiological wall motion movement with a peak displacement of 689 microm. The accuracy of the TDI method was found to decrease with decreasing wall displacement and increasing wall velocity.

Factors affecting the arterial distension waveform derived from tissue Doppler imaging (TDI): an in vitro study on precision.

An investigation was made into the effect of acquisition parameters on the distension waveform estimated from tissue Doppler imaging (TDI). Physiological distension waveforms were generated using a compliant wall phantom. Distensions derived over a range of scanning geometries and transducer pressures were compared with those obtained in optimised scanning conditions. The estimated maximum distension decreased with scanning depth by 7% between 20 mm and 44 mm below the phantom surface, with an increase in transducer-vessel angle (by 22% from 0 degrees to 24 degrees ) and with a decrease in scan plane-vessel coincidence (by 34% from coincidence to 2 mm from the vessel central axis). An increase with transducer pressure was observed (by 20% from contact to high exerted pressure).