Development of a lesion localisation tool to improve outcome prediction in Traumatic Brain Injury patients

Tese de mestrado integrado, Engenharia Biomédica e Biofísica (Engenharia Clínica e Instrumentação Médica) Universidade de Lisboa, Faculdade de Ciências, 2022Traumatic brain injury (TBI) is a highly heterogeneous pathology that poses severe health and socioeconomic problems on a global scale. Neuroimaging research and development has advanced its clinical care in numerous ways, as injured brains are being imaged and studied in greater detail. The size and location of TBI lesions are often necessary to accurately determine a prognosis, which is key in defining a patient-specific rehabilitation program. This dissertation aims to investigate the impact of lesion characteristics, such as volume and location, on outcome prediction in TBI patients. Lesion localisation was achieved by comparing annotated TBI lesions to a brain atlas. Furthermore, other lesion characteristics were examined across different Magnetic Resonance Imaging (MRI) sequences and scanners, with results suggesting that the use of different scanners or MRI contrasts introduced biases in said lesion characteristics. Patient outcome was predicted using four generalised linear models. Besides clinical variables, these models included lesion volume, group and location as predictors. Model comparison indicated that lesion volume could be beneficial for outcome prediction, but may be dependent on both lesion group and location. Overall, this methodology showed potential in uncovering the effect that certain lesion groups and/or locations have on patient outcome after TBI