Cerebrovascular disease, in particular intracranial atherosclerotic stenosis and cerebral aneurysm, is a common cause of stroke, which is the third leading cause of death in many places including Hong Kong. Understanding the geometric char-acteristics of those lesions is an essential part of diagnosis and endovascular treat-ment planning. In the clinical environment, quantitation of the lesions is often conducted on 2-D digital subtraction angiography (DSA). The quantitation may suffer from the foreshortening effect due to inappropriate X-ray projection angle selection and the non-linear magnification caused by the perspective projection. The technological advances in 3-D medical imaging open up new opportunities for the elucidation of the lesion characteristics. This motivates us to de...[ Read more ]

Cerebrovascular disease, in particular intracranial atherosclerotic stenosis and cerebral aneurysm, is a common cause of stroke, which is the third leading cause of death in many places including Hong Kong. Understanding the geometric char-acteristics of those lesions is an essential part of diagnosis and endovascular treat-ment planning. In the clinical environment, quantitation of the lesions is often conducted on 2-D digital subtraction angiography (DSA). The quantitation may suffer from the foreshortening effect due to inappropriate X-ray projection angle selection and the non-linear magnification caused by the perspective projection. The technological advances in 3-D medical imaging open up new opportunities for the elucidation of the lesion characteristics. This motivates us to develop three novel algorithms for the segmentation of vessels and vascular lesions in 3-D angiography.

The first and the second algorithms are low-level segmentation algorithms aiming at producing angiographic segmentation for the third algorithm to per-form high-level segmentation of the atherosclerotic plaque volume and aneurysm body from the diseased portion of the arteries. Such an approach enables later in-teractive or automatic quantitative analyses on the pathological structures. In an in vivo study with nearly twenty angiograms, the results show that our method can help increase the measurement repeatability of the clinical parameters by more than ten times. Therefore, we believe that the works presented can help elucidate the lesion characteristics for diagnosis of the cerebrovascular diseases and endovascular treatment planning. Lastly, the direction of future studies is discussed in this dissertation.