Development of Automatic Detection Methods for Exploration of Lesional and Non-lesional Neurological Disorders with Magnetic Resonance Imaging

• Main Authors: Syu-Jyun Peng, 彭徐鈞
• Other Authors: Jang-Zern Tsai
• Format: Others
• Language: en_US
• Published: 2014
• Online Access: http://ndltd.ncl.edu.tw/handle/94974907317903714887

博士 === 國立中央大學 === 電機工程學系 === 102 === Magnetic Resonance Image (MRI) contains pathology-related information. Detection of MRI-based biomarkers is of diagnostic and therapeutic value. Accurate detection of this type of markers is challenging because they may not be directly discernible and some are even non-lesional on conventional MRI. This dissertation presents different methods for finding lesional biomarkers of acute ischemic stroke and non-lesional ones of neocortical seizures. For lesional biomarkers of acute ischemic stroke, we proposed a computer-assisted segmentation and quantification method to depict cerebral infarct and white matter hyperintensities (WMH). The cerebral infarct and WMH volume were measured based on the histographic distribution of lesions to define self-adjusted intensity thresholds using multispectral MRI. The proposed method attained high agreement with the semi-automatic method. For non-lesional biomarkers of neocortical epilepsy, a popular fiber-labeled MRI template was transformed to each subject’s neuroanatomy to generate personalized atlases for objective and automatic region-of-interest (ROI) demarcation. We investigated supratentorial white matter and subcortical gray matter structures from high-resolution raw structural images and diffusion tensor images with automatic ROI registrations in neocortical seizures. The automatic methods presented in this dissertation facilitates the exploration of lesional and non-lesional biomarkers of neurological disorders for assisting the clinical diagnosis, identifying the risk, and helping guide the treatment and prognosis of the diseases.