Study On The Classification Of Prostate Cancer With Hybrid T2 And Diffusion-Weighted MRI

With the increasing morbidity and mortality rates of prostate cancer(PCa)in recent years,multi-model magnetic resonance imaging(MRI)has become more widely used in the diagnosis and localization of PCa as an accurate and noninvasive way of examination.Research findings and clinical experience show that T2-Weighted Imaging(T2WI)can clearly present the anatomic structure and Diffusion-Weighted Imaging(DWI)can accurately locate the lesion.This thesis used a hybrid T2-DW MRI sequence,which combines the advantages of both T2WI and DWI,to study the key issues in the classification of PCa tissue and evaluate the accuracy of the classification.The main contributions of the thesis include the following two aspects.For the hybrid T2-DW MRI sequence,the thesis developed a PCa tissue classification method based on the Region of Interest(ROI).This method calculated the Apparent Diffusion Coefficient(ADC)and T2 of each voxel,selected its average value in ROI and other statistics as the input,and designed classifiers to study PCa tissue classification.Research shows that Linear Discriminant Analysis(LDA)classifier,whose outputs yielded the area under the receiver operating characteristic curve(AUC)of 0.96 ± 0.03,was more effective in classifying normal and PCa ROIs.The Quadratic Discriminant Analysis(QDA)classifier,whose outputs yielded an AUC of 0.98 ± 0.02,was also effective in classifying benign prostatic hyperplasia(BPH)and PCa ROIs.In the above traditional method,T2WI and DWI were analyzed independently for tissue classification.A Tissue Component Model was proposed for hybrid analysis of the two modalities simultaneously.Based on histopathological observations,the model assumed that only three tissue components(stroma,epithelium and lumen)were present and linearly contributed to the total MR signals.The tissue volume,ADC and T2 were selected as input of the classifiers to differentiate PCa tissue.The model,yielding the maximum AUC of 0.97 ± 0.02,was more effective in differentiating normal and PCa ROIs while similar to traditional method in differentiating BPH and PCa ROIs.With the results in the differentiation of normal and cancer ROIs,benign prostatic hyperplasia and cancer ROIs with the proposed T2-DW modeling method,the thesis shed light on the increase in diagnostic accuracy of PCa.