Algoritmos para detecção de estresse por meio de sensores vestíveis em cenário de vida real: uma revisão sistemática

Abstract

The definition of stress has evolved as researchers and clinicians gain a deeper understanding of the consequences and outcomes it generates for each individual. Recurring exposure to stressors can lead to physical, psychological, and neuropsychiatric consequences, such as metabolic, cardiac, and sleep disorders, as well as executive dysfunctions, most commonly exemplified by depression and anxiety. As a recurring condition associated with biopsychosocial harm, the reliable detection of stress can be an important step in assessing patient health, promoting preventive and personalized medicine. In an effort to make stress detection more objective, new technologies and methodological solutions are being developed and utilized. One such example is the use of wearable devices, which can be easily acquired and used daily by individuals, providing a large amount of information based on physiological parameters. While there are studies analyzing the accuracy of machine learning models for stress detection, few examine this metric in datasets developed from real-life scenario data collections. Therefore, the aim of this systematic review is to identify machine learning models available in the literature for stress detection captured by wearable devices in real-life settings. The PRISMA 2020 Statement checklist was used to ensure good methodological quality in the inclusion of articles for this systematic review. Regarding sample characterization, the size and profile used in each study were divergent. According to the findings, wearable device sensors were used either separately or in combination. In total, 17 model algorithms were utilized. The comprehensive identification of algorithms for stress detection via wearable devices allows us to conclude that this is an important tool for monitoring stressors for each individual and enabling self-management of the condition. Gaps in methodology were also identified, which still create uncertainty regarding the use of wearable devices as a clinical aid tool for healthcare professionals.