Computer Aided Diagnosis Of Soft Tissue Tumors Based On MRI Images

Objective: This study aims to explore MRI-based radiomics on distinguishing malignant from benign soft-tissue tumors.Methods: A set of 85 patients underwent both T1 WI and T1 C MRI sequences(34 benign and 51 malignant)were enrolled in this study.The experiment is divided into three stages,which are introduced in chapter 3,chapter 4 and chapter 5.In chapter 3,support vector machine and k-nearest neighbor were used to predict benign and malignant soft tissue tumors,and an ensemble classifier was constructed by combining two MRI sequences.In chapter 4,k-means unsupervised clustering was used to segment the whole tumor region based on gray value,and radiomics model was constructed based on the handcrafted features.In chapter 5,a three-step method was used to segment the whole tumor region based on gray and entropy values,and Logistic regression was used to construct an radiomics model to draw a nomogram for clinicians.The performance of models was evaluated by the receiver operating characteristic analysis(ROC).The calibration curve and decision curve were used to evaluate the nomogram.Results: In chapter 3,the area under the ROC curve(AUC)of ensemble classifier were0.833 in the training cohort and 0.887 in the validation cohort.In chapter 4,the results showed that the active region/T1 C and the active region/T1 WI were the most discriminative regions,and AUC of the fusion radiomics in the training cohort and the validation cohort were 0.865 and 0.856,respectively.The results of chapter 5 showed that the subregion＿2/T1 C and the whole tumor region/T1 WI were the most discriminative regions,and a fusion radiomics was established with AUC of 0.923 in the training cohort and 0.903 in the validation cohort.Conclusion: Our findings revealed that the three-step intratumor partitioning can better solve the tumor heterogeneity.The diagnostic value of subregion/T1 C sequence was higher than that of T1 WI sequence.The multi-MRI radiomics constructed in this study with potential clinical value can also improve the application of computer-aided diagnosis on differentiation between benign and malignant soft tissue tumors.