| Compared with traditional Computed Tomography(CT)imaging technology,Magnetic Resonance Imaging(MRI)can provide more abundant information.This thesis is based on the MRI images of rectal cancer aided diagnosis and prognostic evaluation.All images come from the Radiation Department of Sichuan people’s hospital.The main work consists the following items:1)The pretreatment of rectal cancer MRI images is studied in this work.The batch conversion of DICOM format file is realized.The pre-processing algorithms of rectal cancer MRI images include: wavelet transform based contrast enhancement algorithm,local contrast enhancement algorithm,gray-level grouping based image contrast enhancement algorithm.These algorithms significantly improve the contrast of abdominal MRI images.2)The location algorithm of rectal region is studied.Based on the analysis of the structure of human pelvic,this thesis proposes an algorithm for rectal region location based on prior information,which uses MRI T2-weighted imaging(T2WI)images of rectal cancer to locate the rectum accurately,and matches the rectum in MRI T2WI sequence images according to the perception hash.3)The segmentation algorithm of rectum and cancer is studied.For the segmentation of rectum and cancer,simple linear iterative clustering(SLIC)combined with fuzzy c-means(FCM)is proposed to segment the rectum accurately on MRI T2WI images,and the automatic region growing method is used to segment the cancer on diffusion weighted imaging(DWI)images.4)The classification algorithm of rectal cancer is studied.Haralick texture feature model,Gabor texture feature model and Local Binary Pattern(LBP)texture feature model are used to extract the features of rectal cancer.Aiming at the problem of redundant features in LBP,Principal Component Analysis(PCA)is used to reduce the dimension of LBP.The fusion of these three features is studied and analyzed.Then support vector machine(SVM)classifier is used to classify the rectal cancer MRI images,and the results are evaluated and analyzed.The accuracy of classification is 91.25%,and the AUC is 0.9034.5)The response evaluation algorithm of rectal cancer is studied.The staging standard of response evaluation is introduced in detail,and the classification is optimized according to the existing data.An algorithm for evaluating the curative effect of rectal cancer is proposed,which combines MRI T2WI and DWI.The preliminary estimation of the curative effect of rectal cancer is completed.The accuracy of the algorithm is 89.66%. |
