Related Experiment Video
Updated: Mar 31, 2026
A Multimodal Imaging Framework to Advance Phenotyping of Living Label-free Breast Cancer Cells
Published on: August 22, 2025
A Multimodal Ultrasound Radiomics Study for Discriminating Benign and Malignant Breast Lesions Classified as BI-RADS
Dongxu Li1, Shanshan Chang1, Chongyang Sun2
1Department of Ultrasound, First Affiliated Hospital of Harbin Medical University, Harbin, Heilongjiang, 150001, People's Republic of China.
Purpose:
This study evaluates multimodal radiomics feature fusion integrating intratumoral and peritumoral radiomics derived from B-mode ultrasound (US) and Virtual Touch Tissue Imaging and Quantification (VTIQ) to differentiate benign from malignant BI-RADS category 4 breast lesions.
Patients And Methods:
A retrospective analysis was conducted on 264 patients. Radiomic features were extracted from intratumoral and peritumoral regions (1 mm, 3 mm, 5 mm) and VTIQ elastography. Three fusion strategies were compared: (i) B-mode US + peritumoral features, (ii) B-mode US + VTIQ, and (iii) triple fusion (B-mode US + peritumoral + VTIQ). Features were selected via LASSO regression, and models (LR, SVM, RF) were developed and validated on separate training (n = 184) and test (n = 80) cohorts.
Results:
The B-mode US model achieved area under the curve (AUC) values of 0.890 (training) and 0.844 (test), surpassing the VTIQ model (training AUC = 0.850, test AUC = 0.812). Fusion of B-mode US and VTIQ features improved performance to AUCs of 0.913 and 0.874, respectively. Among peritumoral margins, the 3 mm region provided the best discrimination, with AUCs of 0.893 (training) and 0.871 (test). Integration of B-mode US with peritumoral features further improved performance (AUC = 0.908 training, 0.859 test). The triple-fusion model (B-mode US + peritumoral + VTIQ) achieved the highest diagnostic accuracy, with AUCs of 0.933 in the training cohort and 0.903 in the test cohort.
Conclusion:
Multimodal feature fusion significantly enhances the differentiation of benign and malignant breast lesions, supporting its potential for clinical decision-making.
