Quantitative Identification of Nonmuscle‐Invasive and Muscle‐Invasive Bladder Carcinomas: A Multiparametric MRI Radiomics Analysis

Abstract

Background Preoperative discrimination between nonmuscle‐invasive bladder carcinomas (NMIBC) and the muscle‐invasive ones (MIBC) is very crucial in the management of patients with bladder cancer (BC). Purpose To evaluate the discriminative performance of multiparametric MRI radiomics features for precise differentiation of NMIBC from MIBC, preoperatively. Study Type Retrospective, radiomics. Population Fifty‐four patients with postoperative pathologically proven BC lesions (24 in NMIBC and 30 in MIBC groups) were included. Field Strength/Sequence 3.0T MRI/T2‐weighted (T2W) and multi‐b‐value diffusion‐weighted (DW) sequences. Assessment A total of 1104 radiomics features were extracted from carcinomatous regions of interest on T2W and DW images, and the apparent diffusion coefficient maps. Support vector machine with recursive feature elimination (SVM‐RFE) and synthetic minority oversampling technique (SMOTE) were used to construct an optimal discriminative model, and its performance was evaluated and compared with that of using visual diagnoses by experts. Statistical Tests Chi‐square test and Student's t‐test were applied on clinical characteristics to analyze the significant differences between patient groups. Results Of the 1104 features, an optimal subset involving 19 features was selected from T2W and DW sequences, which outperformed the other two subsets selected from T2W or DW sequence in muscle invasion discrimination. The best performance for the differentiation task was achieved by the SVM‐RFE+SMOTE classifier, with averaged sensitivity, specificity, accuracy, and area under the curve of receiver operating characteristic of 92.60%, 100%, 96.30%, and 0.9857, respectively, which outperformed the diagnostic accuracy by experts. Data Conclusion The proposed radiomics approach has potential for the accurate differentiation of muscle invasion in BC, preoperatively. The optimal feature subset selected from multiparametric MR images demonstrated better performance in identifying muscle invasiveness when compared with that from T2W sequence or DW sequence only. Level of Evidence: 3 Technical Efficacy: Stage 2 J. Magn. Reson. Imaging 2019;49:1489–1498.

A predictive nomogram for individualized recurrence stratification of bladder cancer using multiparametric MRI and clinical risk factors

“…MpMRI including T 2 ‐weighted (T 2 W), DW, and DCE image sequences were performed to obtain the T 2 W, DW, and DCE images. The corresponding apparent diffusion coefficient (ADC) maps were generated automatically from the DW images . Primary parameters of these sequences were described in Appendix E1.…”

“…Preoperative radiomics features, such as the histogram features, Haralick features extracted from co‐occurrence matrix (CM), and features extracted from run‐length matrix (RLM), neighborhood gray‐tone difference matrix (NGTDM), and gray level size zone matrix (GLSZM), have demonstrated their potential in recurrence risk stratification of many other cancer diseases . In this study, radiomics features, including eight histogram features, 39 CM features, 33 RLM features, five NGTDM features, and 15 GLSZM features, were extracted from tumor ROI of each image modality to fully characterize the local, regional, and global tissue distribution variations of the tumor . Detailed information about these features was shown in Table E1 of Appendix E1.…”

“…Concerning different grayscales of the original images acquired, prior to CM, RLM, NGTDM, and GLSZM features extraction, the ROI grayscales were discretized and normalized to multiple grayscales, including 8, 16, 32, 64, and 128 gray level, respectively. For each tumor ROI, 468 features including eight histogram features, 195 CM features (39 × 5), 165 RLM features (33 × 5), 25 NGTDM features (5 × 5), and 75 GLSZM features (15 × 5) were calculated.…”

“…During each sequential iteration, the feature that has the least impact on improving the classification performance is omitted. As the algorithm continues, all the features are removed one by one, and the feature ranking list is generated based on the number of iterations when the feature was discarded . In the end, the top‐ranked features are those that remained the last in the iterations.…”

“…In the end, the top‐ranked features are those that remained the last in the iterations. Based on the ranking list, the optimal subset that contains the first n features and obtains the highest AUC in classification is finally determined …”