| Summary: | 碩士 === 義守大學 === 資訊工程學系碩士班 === 98 === For moving into aging society, Parkinson''s disease has become one of the most common ailments around the world. In the primary stage, the symptom is often not so clear for those patients with Parkinson''s disease. Although it will not immediately result in death, but it is a heavy burden for the family and society. In order to achieve the objective of early detection and early treatment, the aim of this thesis is to develop an automatic 3D image segmentation and 3D geometric features of TRODAT-1 images fitted with dopamine''s activity for providing physician in the diagnosis of Parkinson''s disease. First, setting the threshold of 75% of maximum intensity is for rough image segmentation, and the 3D connected component labeling is used to remove noise for retaining two main striatum objects. Then, some geometric invariant features, including volume and 3D shape histograms, are proposed to extract the 3D shape characteristics of segmented objects. Based on these features, the volumes and the similarity of left and right segmented striatum can be estimated so as to predict the dopamine’s activity. Finally, ROC curves are used as an accuracy test for the diagnosis of Parkinson''s disease.
|
|---|
