A Machine Learning‐Based Unenhanced Radiomics Approach to Distinguishing Between Benign and Malignant Breast Lesions Using T2‐Weighted and Diffusion‐Weighted MRI

• DOI: 10.1002/jmri.29111
• Publication Date: 2023-11-07T13:30:08Z
• Authors: Yulu Liu; Xiaoxuan Jia; Jiaqi Zhao; Yuan Peng; Xun Yao; Xuege Hu; Jingjing Cui; Haoquan Chen; Xiufeng Chen; Jing Wu; Nan Hong; Shu Wang; Yi Wang
• Keywords: Breast MRI; Breast imaging; Kappa; BI-RADS

Background Breast MRI has been recommended as supplemental screening tool to mammography and breast ultrasound of cancer by international guidelines, but its long examination time use contrast material remains concerning. Purpose To develop an unenhanced radiomics model with using non‐gadolinium based sequences for detecting on T2‐weighted (T2W) diffusion‐weighted (DW) MRI. Study Type Retrospective analysis followed retrospective prospective cohorts study. Population 1760 patients: Of these, 1293 construction ( n = 775 training 518 validation). The remaining patients testing in internal 167), 188), external 112) cohorts. Field Strength/Sequence 3.0T MR scanners from two institution. T2WI, DWI, first contrast‐enhanced T1‐weighted sequence. Assessment AUCs distinguishing were compared between combined gadolinium agent sequence model. Subsequently, the was radiologists' diagnosis this research. Finally, patient subgroup performed clinical subgroups different types malignancies. Statistical Tests Mann–Whitney U test, Kruskal‐Wallis H chi‐square weighted kappa DeLong's test. Results best under Gaussian process (GP) classifiers (AUC: training, 0.893; validation, 0.848) support vector machine (SVM) logistic, showing favorable prediction (AUCs, 0.818–0.840). not statistically five those P , 0.317–0.816), well radiologists 0.181–0.918). least successful identifying ductal carcinoma situ, whereas did show statistical significance other subgroups. Data Conclusion An T2WI DWI comparable diagnostic accuracy agent. Level Evidence 4 Technical Efficacy Stage 2