Identifying post-COVID-19 syndrome neuropsychiatric subgroups through resting-state fMRI and graph-based machine learning

• Grant number: 23/11469-1
• Support Opportunities: Scholarships abroad - Research Internship - Doctorate (Direct)
• Effective date (Start): January 01, 2024
• Effective date (End): June 30, 2024
• Field of knowledge: Health Sciences - Medicine - Medical Clinics
• Principal Investigator: Clarissa Lin Yasuda
• Grantee: Ítalo Karmann Aventurato
• Supervisor: David Kohan Marzagão
• Host Institution: Faculdade de Ciências Médicas (FCM). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil
• Research place: King's College London, England
• Associated to the scholarship: 22/11786-4 - Long COVID: phenotyping and characterization of the functional and structural connectivity in patients with cognitive sequelae from COVID-19, BP.DD

Abstract

The post-COVID syndrome may present even after mild and moderate illness with an incidence of up to 20\%, leading to significant functional impairment of those inflicted. Neuropsychiatric symptoms are one of the main components of this syndrome, presenting with fatigue, memory impairment and executive dysfunction. In this context, describing neuropsychological subtypes of the post-COVID syndrome is paramount to better understanding this disease. Additionally, characterizing the structural and functional brain changes associated with these subgroups may add to the comprehension of these phenotypes and their pathophysiology. Machine learning techniques in neuroimaging data are currently on the rise. The emergence of methods capable of dealing with graph-structured data, such as graph kernels and graph neural networks, allows the application of these methods to the results of resting-state fMRI brain connectivity analysis. In this project, we aim to study these methods with neuroimaging data in two steps: first, in a pilot study with data from patients with temporal lobe epilepsy seeking to localize the epileptogenic zone and, subsequently, with data from the UNICAMP's NeuroCOVID cohort in an attempt to identify COVID-19 infected subjects from controls as well as their neuropsychological subtypes. The project will be developed in collaboration with the King's College London computational sciences department along with their expertise in the study of graphs and machine learning applied to many different scientific problems. The work developed until now in the context of this doctoral thesis has shown the validity of the neuropsychological constructs used and was capable of identifying neuropsychological subtypes among those previously infected by COVID-19. Furthermore, preliminary work with a simple classifier (Support Vector Classifier) has shown that these techniques can potentially identify connectivity abnormalities in those infected by COVID-19. (AU)