Artificial Intelligence Analysis for Magnetic Resonance Imaging in Screening and Diagnosis of Breast Cancer

Peking University People's Hospital Radiology

Use Convolutional Neural Networks Analysis for Classification of Contrast-enhancing Lesions at Multiparametric Breast MRI. Build an abbreviated protocal, and investigate whether an abbreviated protocol was suitable for breast magnetic resonance imaging screening for breast mass in Chinese women, which can shorten the examination time and avoid enhanced imaging while ensuring the accuracy of the diagnosis.

Inclusion Criteria:

• Patients with clinical symptoms (define as palpable mass, nipple discharge, asymmetric thickening or nodules, and abnormal skin changes)
• Patients undergoing full sequence BMRI examination
• Through the follow-up database, at least 6 months of follow-up results can be obtained to determine whether the diagnosis result is negative/benign/malignant; for patients who need pathological biopsy, the pathological biopsy results shall prevail to determine the lesion benign/malignant

Exclusion Criteria:

• The breast had received radiotherapy, chemotherapy, biology and other treatments before BMRI.
• There are contraindications for breast-enhanced MRI examinations such as allergy to contrast agents.
• A prosthesis is implanted in the affected breast.
• Patients during lactation or pregnancy