Differentiation of benign and malignant breast lesions with a platform pattern time-signal intensity curve:a radiomics model based on dynamic contrast-enhanced and diffusion-weighted imaging
YANG Ke
WANG Xuzhao
LI Yue
HUANG Jiarong
MIAO Chongchang
Abstract：Objective To evaluate the diagnostic value of radiomics model based on dynamic contrast-enhanced(DCE)and diffusion-weighted imaging(DWI)sequences in differentiating benign from malignant breast lesions with platform-type time signal intensity curve(TIC).Methods A retrospective analysis was conducted involving a cohort of 251 patients who underwent breast DCE at the Lianyungang Clinical Medical College of Nanjing Medical University from January 1st,2019 to October 30th,2023.All diagnoses were confirmed by pathological examination.TIC curves were generated for all cases utilizing manufacturer image post-processing workstation,resulting in a total of 113 cases with platform-type TIC.This dataset comprises 78 instances of malignant breast lesions and 35 instances of benign lesions.All lesions are randomly allocated into training and testing datasets in a ratio of 7:3.The region of interest of each lesion was manually segmented,and support vector machine(SVM)classifier was employed to develop radiomics models.The ROC curve was constructed,and the area under the ROC curve(AUC),sensitivity,and specificity were computed to assess the differential diagnostic efficacy of the models.Results A total of 1502 features were individually extracted from each sequence.After dimensionality reduction screening using t-test and Least Absolute Shrinkage and Selection Operator(LASSO)regression,9 features are selected from the DWI sequence,6 from the DCE sequence,and 11 from the DWI-DCE sequence.For DWI model training dataset,the AUC was 0.77 with a sensitivity of 84%and a specificity of 96%;for the testing dataset it was 0.72,81%and 96%.For DCE model training dataset it was 0.87,90%and 98%;for testing dataset it was 0.76,84%and 98%;For DWI-DCE model training dataset it was 0.84,88%and 96%;for testing dataset it was 0.75,81%and 96%.Conclusion Compared to the DWI model,both the DCE model and the combined DCE-DWI model shows better performance in distinguishing benign and malignant breast lesions with platform-type TICs.
Keywords：dynamic enhanced magnetic resonance imagingtime signal intensity curvebreast cancerradiomics
Publication Date：2026-02-20
Online Publishing Date：2026-03-25(First online date of this platform, not the publication date of the document)
Pages：7( 231-237 )