# Use of physiological signals, behavioral data, and processing algorithms in electronic devices and mobile applications for diagnosing depression, anxiety, and stress

## TL;DR

This paper reviews how electronic devices and apps use physiological and behavioral data to detect depression, anxiety, and stress, highlighting their potential and current limitations.

## Contribution

A systematic review of 77 studies on technologies and algorithms for diagnosing mental disorders using wearable and mobile tools.

## Key findings

- Wrist-based devices, like smartwatches, are most commonly used for monitoring heart rate and sleep patterns.
- Classification algorithms achieved 75-90% accuracy in diagnosing mental disorders.
- Challenges include generalizability and the need for personalized models in real-world settings.

## Abstract

To review the technologies, biomarkers and processing algorithms used in diagnosing depression, anxiety, and stress, informing future research endeavors to enhance healthcare and the well-being of individuals affected by these mental disorders.

A systematic review was conducted through searches in electronic databases such as PubMed, Google Scholar, IEEE, Nature, ProQuest, and Science Direct. Search queries combined terms related to physiological signals, behavioral variables, electronic devices, mobile applications, and the disorders of depression, anxiety, and stress. After screening 292 initial records, 77 studies met specific criteria, which included discussion of quantitative results and advanced processing algorithms.

The review of 77 articles revealed an increasing use of electronic devices and applications for measuring physiological and behavioral variables in the diagnosis of mental disorders. The wrist was the most common device location, accounting for 53.2%, primarily utilizing smartwatches to monitor heart rate, electrodermal activity, and sleep patterns. The analyzed technologies included wearable sensors, facial recognition cameras, electroencephalographs, and virtual reality devices. The classification algorithms used—such as decision trees, neural networks, and support vector machines—achieved accuracy rates ranging from 75% to 90%, highlighting the effectiveness of these tools. However, limitations were identified regarding the generalizability of results and the need for personalized diagnostic models.

Electronic devices and mobile applications represent a significant advancement in the detection and monitoring of depression, anxiety, and stress by providing objective data continuously and in real time. However, their clinical application still faces challenges related to accuracy, personalization, and user acceptance. For these technologies to be effectively integrated into clinical practice, it is recommended to conduct studies in real-world settings and foster collaboration with mental health professionals. Such efforts would enable the adaptation of diagnostic models to individual needs and enhance the accuracy of early interventions.