Dissertação de Mestrado

Generative AI and Physiological Sensors for Obsessive Compulsive Disorder Exposure Therapy

Maria Beatriz Dias Correia2025

Informações chave

Autores:

Maria Beatriz Dias Correia (Maria Beatriz Dias Correia)

Orientadores:

Hugo Humberto Plácido da Silva (Hugo Humberto Plácido da Silva); Alexander Loktyushin

Publicado em

24 de novembro de 2025

Resumo

Exposure and Response Prevention (ERP) is the gold-standard therapy for Obsessive-Compulsive Disorder (OCD), a debilitating anxiety disorder. This treatment approach faces challenges related to personalization and long-term effectiveness. This thesis investigated the feasibility and preliminary efficacy of a novel system designed to enhance exposure therapy for contamination-subtype OCD by integrating generative Artificial Intelligence (AI) imagery and physiological sensors (Electrodermal Activity (EDA), Electrocardiogram (ECG), and accelerometer). A proof-of-concept was developed to determine whether exposure to AI-generated imagery can provoke a measurable emotional response. A complete signal processing pipeline was implemented and relevant features were extracted. Its performance was validated through in silico simulations reproducing expected autonomic responses under experimental conditions. Real-patient results showed no statistically significant differences across physiological metrics when comparing exposure and neutral conditions. Therefore, this static, two-dimensional generative AI implementation is likely insufficient to induce the necessary emotional and physiological responses for effective therapy outcomes. Participant feedback supported these findings, with common concerns including limited realism, weak immersion, and lack of image personalization. While the multimodal acquisition and analysis framework proved technically feasible, the quality of the stimulus delivery remains a major barrier. Findings highlight the need for increased immersion, possibly in a Virtual Reality (VR) environment. This work constitutes a strong foundation for future research toward real-time, biofeedback-based adaptive systems for OCD exposure therapy.

Detalhes da publicação

Autores da comunidade :

Orientadores desta instituição:

Domínio Científico (FOS)

industrial-biotechnology - Biotecnologia Industrial

Idioma da publicação (código ISO)

eng - Inglês

Acesso à publicação:

Acesso Embargado

Data do fim do embargo:

23 de outubro de 2026

Nome da instituição

Instituto Superior Técnico