Adoption of Big Data and AI methods to manage medication administration and intensive care environments

• Autor(a) principal: Oliveira, João Pedro Alves
• Data de Publicação: 2023
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10362/155445
• topic: Telemedicine; Artificial Intelligence; Machine Learning; Decision Support Systems; Domínio/Área Científica::Engenharia e Tecnologia::Engenharia Eletrotécnica, Eletrónica e Informática

Artificial Intelligence (AI) has proven to be very helpful in different areas, including the medical field. One important parameter for healthcare professionals’ decision-making process is blood pressure, specifically mean arterial pressure (MAP). The application of AI in medicine, more specifically in Intensive Care Units (ICU) has the potential to improve the efficiency of healthcare and boost telemedicine operations with access to real-time predictions from remote locations. Operations that once required the presence of a healthcare professional, can be done at a distance, which facing the recent COVID-19 pandemic, proved to be crucial. This dissertation presents a solution to develop an AI system capable of accurately predicting MAP values. Many ICU patients suffer from sepsis or septic shock, and they can be identified by the need for vasopressors, such as noradrenaline, to keep their MAP above 65 mm Hg. The presented solution facilitates early interventions, thereby minimising the risk to patients. The current study reviews various machine learning (ML) models, training them to predict MAP values. One of the challenges is to see how the different models behave during their training process and choose the most promising one to test in a controlled environment. The dataset used to train the models contains identical data to the one generated by bedside monitors, which ensures that the models’ predictions align with real-world scenarios. The medical data generated is processed by a separate component that performs data cleaning, after which is directed to the application responsible for loading, classifying the data and utilising the ML model. To increase trust between healthcare professionals and the system to be developed, it is also intended to provide insights into how the results are achieved. The solution was integrated, for validation, with one of the telemedicine hubs deployed by the European project ICU4Covid through its CPS4TIC component.