Design and Evaluation of a Multimodal Medical Device-Based Framework for COVID-19 Inference and Risk Classification Based on the Manchester Triage System
Name: LETICIA ARAÚJO SILVA
Publication date: 15/08/2025
Examining board:
Name |
Role |
|---|---|
| ADRIANO DE OLIVEIRA ANDRADE | Examinador Externo |
| ANA CECILIA VILLA PARRA | Examinador Externo |
| DENIS DELISLE RODRÍGUEZ | Examinador Externo |
| EDUARDO LAZARO MARTINS NAVES | Examinador Externo |
| ELIETE MARIA DE OLIVEIRA CALDEIRA | Examinador Externo |
Pages
Summary: The rapid and accurate classification of patients in emergency contexts is essential to optimize clinical decision-making and resource allocation. This dissertation presents a multimodal signal-based framework for intelligent medical triage, encompassing two complementary applications: (i) the automatic detection of COVID-19 and (ii) the classification of clinical risk levels according to the Manchester Triage System. Both tasks rely on the analysis of respiratory sounds and vital signs, acquired from benchmark datasets and validated using the Integrated Portable Medical Assistant (IPMA), a portable medical device designed for multimodal data collection. To support inference, different preprocessing and feature extraction strategies were employed, followed by the training and evaluation of machine learning models, including deep neural networks. The models were benchmarked using publicly available datasets and subsequently tested with real-world data collected through the AMPI device. Performance was assessed using metrics such as accuracy and F1-score. Statistical tests were applied to compare classifiers and validate improvements. Results demonstrated high performance in both COVID-19 detection and Manchester-based risk classification, with models achieving competitive accuracy and robustness even when trained with limited and heterogeneous data. Usability analysis, based on the System Usability Scale (SUS) and Post Study System Usability Questionnaire (PSSUQ), indicated strong acceptance and user satisfaction with the system interface. The proposed framework reinforces the feasibility of machine-assisted triage using low-cost and portable solutions, especially in settings with limited access to healthcare infrastructure.
