Development of an AI System for Smart Safe Health Monitoring

• Autor(a) principal: Barbosa, Luís Carlos Novais
• Data de Publicação: 2022
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/11110/2545
• topic: Artificial Intelligence; Health; Monitoring; Magnet Marker Tracking; Biosignals

In this document, the application of Artificial Intelligence (AI) in Health is being investigated in medical procedures and monitoring strategies. Regarding the first, a strategy based on Magnet Tracking (MT) was developed, which helps the common procedure for the treatment of fractures of the tibial shaft, where an intramedullary nail (IN) is placed through the medullary canal of the tibia and respective distal and proximal interlocks (PI and DI) in a way that to fix the same. The navigation process of this procedure is performed by real-time two-dimensional medical imaging strategies (fluoroscopy). To avoid excess radiation, it was followed the methodology adopted by the MT technique, together with AI, to accurately estimate the magnetic marker position and consequently the optimal drilling location. The potential of AI techniques for population monitoring is being investigated, for the diagnosis of serious pathological situations through the analysis of vital signs. A study was carried out in the literature to verify which biosignal databases are available for free. A total of thirteen public databases were identified, nine referring to the ECG signal, and the remaining four referring to other biosignals. AI methods show a high performance in their classification purposes. In this sense, a third study was prepared to study the robustness of three state-of-the-art methods, based on deep learning (DL), developed to classify ECG segments, after the introduction of three public ECG signal databases, to verify if the selected methods present classification performance independently of the selected database. In summary, the results report the potential of the device, together with the MT and the AI to estimate the optimal drilling location for placement of the PI and DI, in the treatment of tibial shaft fractures. Then, the study carried out described the public biosignal databases available in the literature, for the creation of intelligent classification methods. The last study allowed us to conclude that only one network architecture achieved high robustness for the variation of the data source, as well as those future studies, are needed to evaluate the performance of this type of AI network in real clinical situations. In general, the presented dissertation reports the potential of AI for application in Health, in two main aspects, in treatment and monitoring.