To develop and compare multi-modality radiomics models based on DCE-MRI, B-mode ultrasound (BMUS) and strain elastography (SE) images for classifying benign and malignant breast lesions.

In this retrospective study,555 breast lesions from 555 patients who underwent DCE-MRI, BMUS and SE examinations were randomly divided into training (n =388) and testing (n = 167) datasets. Radiomics features were extracted from manually contoured images. The inter-class correlation coefficient (ICC), Mann-Whitney U test and the least absolute shrinkage and selection operator (LASSO) regression was applied for feature selection and radiomics signature building.Nine radiomics models including four single-modality radiomics models (DCE-3D, DCE-2D, BMUS, and SE), four multi-modality radiomics models (BMUS + SE, DCE-3D + BMUS, DCE-3D + SE, and DCE-3D+SE+BMUS), and the combination diagnostic model(clinical features and DCE-3D+SE+BMUS features) were developed and evaluated by their discrimination, calibration, and clinical usefulness.

Inclusion Criteria:

Patients with breast lesions underwent biopsy or surgical resection between January 1, 2018 and March 30, 2024.

Exclusion Criteria:

1. pathologicalresult of biopsy or surgical specimen was unavailable for the target lesion;
2. patients without DCE-MRI, BMUS and SE examinations before biopsy or surgery within one month;
3. patients hadperformed radiotherapy, chemotherapy, or breast biopsy before MRI and ultrasound examinations;
4. patients without completeDICOM data for each examination;
5. patients with poor-qualityimages.

Patients with breast lesions attending the First Affiliated Hospital of Shandong First Medical University were selected

Lack of resources or infrastructure: Sharing IPDs requires resources and infrastructure to ensure data security, manage access requests and provide necessary documentation. Currently these resources are not well developed, so it may be difficult to share IPDs.