| Summary: | 碩士 === 國立成功大學 === 資訊工程學系 === 105 === Multiple Sclerosis (MS) is a relatively common inflammatory disease involving the central nervous system. MS lesions vary greatly in shape, location, intensity and area, which challenge the automated segmentation methods. Thus, the lesion load and its delineation have established their importance for assessing disease progression. This work built an automated system includes brain extraction, registration, atlas model creation, bias correction and tissue segmentation with MS lesions and other tissues.
In this work, we further extend Multi-channel MICO (MCMICO) algorithm and modify the energy formulation by introducing the atlas probability model into the MR images. According to the characteristic of MS lesion which primarily affects the white matter, the study enhances MS lesions and also segments other tissues by applying the probability map.
The proposed method is validated by comparing with the original MCMICO algorithm and an existing toolbox LPA. The measures mostly demonstrate a great improvement of our method. With introducing an atlas probability model as the priori knowledge, the segmentation method effectively rejects the false positives. After post-processing, the proposed method further improves the results.
|
|---|
