The Study of Intracranial Hematoma Image Segmentation Method Based on Extension Detecting and Support Vector Machine
|School||Guangdong University of Technology|
|Course||Control Theory and Control Engineering|
|Keywords||Extension detection matter-element transformation support vector machine intracranial hematoma medical image segmentation|
Medical image segmentation is a research hotspot and difficult point in recent years. For different research objects, it needs design algorithm according to detail circumstances, so far there is no universal segmentation algorithm suitablefor all objects. Intracranial hematoma, especially acute intracranial hematoma is still one of disease symptom threatening life-health,. Automatical and accurate segmentation of intracranial hematoma region on CT medical image region, lays solid foundation for three-dimensional reconstruction and volume actuary of intracranial hematoma, furtherly provides better help for doctor’s clinical diagnosis and patients’s understanding of disease state,, and has great clinical application significance.The research emphasis of this subject is automatic segmentation method of intracranial hematoma in medical image.Thesis innovation points are as follows: Firstly, it’s the first time to propose the method of combining matter-element transformation method of extension detecting theory and support vector machine algorithm to achieve segmentation of intracranial hematoma in medical image; Secondly, adopting optimization algorithms to optimize parameters of support vector machine in order to obtain good segmentation result.In this thesis, segmentation process of intracranial hematoma in CT medical image has two steps. The first step is to use threshold segmentation method and region growing method to achieve initial segmentation of intracranial hematoma in CT medical images. The second step is to use support vector machine (SVM) method to make classifier training with initial segmentation image and hematoma region segmentation. According to the problem that SVM segmentation result exists some points of non-hematoma region, which are regarded as the point in hematoma region and makes unsatisfied segmentation result, this thesis firstly proposes to combine the thought of solving contradiction problem in matter-element transformation method of extension detecting theory to carry on extension optimization for SVM segmentation results. Establishing measurable-element model,, unmeasurable-element model and target-element model for SVM segmentation results. By transforming unmeasurable hematoma region into area-measurable region, obtains the maximal hematoma area.Matlab simulation experimental results show that segmentation algorithm of intracranial hematoma in CT image proposed in this paper is effective, which only needs train one SVM classifier to get automatic and accurate segmentation of intracranial hematoma with the maximal area for entire sequence of one patient’s intracranial hematoma in CT images, and the segmentation effect is satisfied requirements of clinical application.