07 May 2024
14:00 Master's Defense IC3 Auditorium
Theme
Uncovering Mental Well-Being: Detecting Stress, Anxious State and Stressors in a Case Study
Student
Matheus Correa Lindino
Advisor / Teacher
Anderson de Rezende Rocha - Co-supervisor: Aurea Rossy Soriano Vargas
Brief summary
Stress and anxiety are natural defense mechanisms of the human body. A stress response can be defined as a state of emergency in the body in response to an internal or external stimulus, which unbalances its homeostasis. In turn, anxiety is defined as a temporally diffuse emotional state caused by a potentially harmful situation. According to the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5), both stress and anxiety, when left untreated, can develop pathological illnesses, such as post-traumatic stress disorder, generalized anxiety disorder and depression. Generally, identification of these diseases involves interviews with professionals or self-report questionnaires by the individual. However, identification can be highly challenging compared to other conditions that can produce similar symptoms. Therefore, more objective metrics capable of identifying stress and anxiety early, before they harm health, are extremely important, consequently improving the treatment of individuals. Although the relationship between psychological stress and anxiety seems intuitive, the biological nuances that distinguish these two states are complex. In the physiological scope, these two concepts overlap, since they both occur together. In this context, this work aims to discuss the relationship between stressors, stress and state of anxiety, identifying their semantic differences and the physiological impacts of each one, using signs such as heartbeat, galvanic response, blood pressure, among others. To this end, this work presents two convolutional network architectures, one to understand the individual contribution of signals and another proposed for input with N sensors, which accepts signals of different sampling frequencies. Furthermore, it presents a new validation setup based on Leave-One-Subject-Out (LOSO), RepeatedLOSOCV, which aims to show more accurate results, taking into account the intra- and interbiological differences of different users.
Examination Board
Headlines:
| Anderson de Rezende Rocha | IC / UNICAMP |
| Agma Juci Machado Traina | ICMC / USP |
| Helena de Almeida Maia | IC / UNICAMP |
Substitutes:
| Paula Dornhofer Paro Costa | FEEC / UNICAMP |
| Ahmed Ali Abdalla Esmin | DCC / UFLA |
| Levy Boccato | FEEC / UNICAMP |