Carotid atherosclerosis is a major risk factor for ischaemic stroke which is a leading cause of death worldwide. For stroke survivors, 1 in 4 will have another stroke within five years. Carotid CT angiography (CTA) is commonly performed following an ischaemic stroke or transient ischemic attack to help guide patient management in the secondary prevention of stroke. For example, carotid endarterectomy surgery plus medical therapy or medical therapy alone. The degree of carotid stenosis is the mainstay in making this decision and uses only one aspect of anatomical information that can be obtained from a carotid CTA scan. Radiomics, sometimes called ‘texture analysis’, is the extraction of quantitative data from medical images that may not be apparent to the naked eye and has already demonstrated clinical utility in oncology for applications ranging from lesion characterisation to tumour grading and prognostication. Machine learning refers to the process of learning from experience (in this case data), rather than following pre-programmed rules. This thesis presents the findings of a proof-of-principle study to assess the value of radiomics in identifying the ‘vulnerable plaque’ and the ‘vulnerable patient’ within the context of cerebrovascular events. To evaluate the potential of radiomic features as imaging biomarkers, their reproducibility and robustness to morphological perturbations were assessed, as well as their biological associations with both PET and immunohistochemistry data. The ability of radiomic features to classify different carotid artery types, namely, culprit, non-culprit and asymptomatic carotid arteries was assessed using several machine learning classifiers. This was subsequently compared with a deep learning approach, which has greater capacity for data mining than feature-based machine learning approaches. Overall, radiomics could extract further useful information from carotid CTA scans. Culprit versus non-culprit carotid arteries in symptomatic patients and asymptomatic carotid arteries from asymptomatic patients had different radiomic profiles that could be leveraged using machine learning for better classification performance than carotid calcification or carotid PET imaging alone. Reliable and robust CT-based carotid radiomic features were identified that were associated with the degree of inflammation underlying the carotid artery. If validated with future prospective studies, this has the potential to improve personalised patient care in stroke management and advance clinical decision-making.