Feasibility Study Of Dominant Intraprostatic Lesions Identification Based On KVCT Image

The prostate is the largest subsidiary of male gonads, belongs to human exocrine. Gene mutation leads to uncontrolled proliferation of prostate cells, and prostate cancer has become. Malignant cells can lead to prostate volume expanded or invade adjacent organs. And it may also be transferred to other parts of the body, especially the bones and lymph nodes. Prostate cancer may cause pain, difficulty urinating, erectile dysfunction and other symptoms. In Western countries, prostate cancer is the second most common cancer in men. The number of deaths due to prostate cancer is second only to lung cancer. Now research on prostate, mainly for reconstruction segmentation of the prostate gland and prostate cancer segmentation in MRI image or ultrasound images. In the CT image, because of the contrast prostate lesions and normal tissue is very low, lesion identification is difficult. Therefore, prostate lesions on CT image identification is almost empty. Compared to common CT images, KVCT images are clear, and the image resolution is high. So, for prostate lesions on KVCT image identification, we have mainly completed the following works:1. Aiming at the problem of low contrast between the lesion and normal tissue in the prostate KVCT image, a method of region recognition based on texture features is proposed. In this method, the two-dimensional gray-gradient feature of the prostate region is extracted firstly. Then the SVM classifier is trained to classify the lesion area. Finally, using a Markov random field model for identification results, obtained the lesion more accurately.2. 3D prostate lesion identification algorithm. In this method, prostate interlayer interpolated firstly. Then three dimensional gray-gradient features are extracted, choosing a certain percentage as training samples. At the same time the three dimensional position information of training sample obtained. Then training SVM, the lesion area based on the SVM was obtained. Then according to the gradient feature and shape constraint, similarity search near the training sample to obtain a lesion area. Then this method uses interactive set level method to determine the extent of the lesion area of the prostate as a constraint, and obtained the fusion result of SVM identification results and the lesion area based on gradient information and shape constraints. Finally, using mathematical morphology methods to remove noise, to get the final three-dimensional prostate lesion area. Experiments show that the method of identifying 3D prostate lesion area has a good effect.