Using Deep Learning and Radiomics to Diagnose Benign and Malignant Breast Lesions Based on Ultrasound

Ultrasound-based Deep Learning Signature and Radiomics Signature Nomogram for Diagnosis of Benign and Malignant Breast Lesions of BI-RADS Category 4 Using Intratumoral and Peritumoral Regions

This retrospective study aimed to create a prediction model using deep learning and radiomics features extracted from intratumoral and peritumoral regions of breast lesions in ultrasound images, to diagnose benign and malignant breast lesions with BI-RADS 4 classification.

Materials and methods: Patients who visited in The First Affiliated Hospital of Shandong First Medical University ＆ Shandong Provincial Qianfoshan Hospital were collected. Their general clinical features, information on preoperative ultrasound diagnosis, and postoperative pathologic data were reviewed.

Inclusion Criteria:

• female patients with US-visible solid breast masses who underwent biopsy and/or surgical resection, and were classified as having BI-RADS 4 lesions in medical US reports.

Exclusion Criteria:

• preoperative endocrine therapy, chemotherapy, or radiotherapy, preoperative invasive breast operation, insufficient image quality, and no pathological results.