Objective To investigate the diagnostic value of an MRI-based deep learning model in breast cancer.

Methods In this cross-sectional study, the clinical and imaging data of 95 patients with breast tumors (female, aged (54.4±8.0) years) who received diagnosis and treatment at Changsha Hospital of Traditional Chinese Medicine from January 2022 to May 2025 were retrospectively analyzed. Among the 95 patients, 95 lesions were found, 31 and 64 of which were histopathologically confirmed to be benign and malignant, respectively. The patients were divided into a training set (63 cases, including 19 benign cases and 44 malignant cases) and a validation set (32 cases, including 12 benign cases and 20 malignant cases) at a ratio of 2∶1 using simple random sampling. All patients underwent MRI examination. An MRI-based deep learning model was constructed for the benign and malignant classification of the lesions. The receiver operating characteristic curve was drawn, and the area under the curve (AUC) was calculated to evaluate the diagnostic performance of the deep learning model for breast cancer. The diagnostic results of the model were compared with those of the senior and junior radiologist groups. DeLong test was used to compare AUCs.

Results In the training set, the AUC of the deep learning model for diagnosing breast cancer was 0.928, which was comparable with that of the senior radiologist group (AUC=0.917) (Z=0.465, P=0.642); both were significantly higher than that of the junior radiologist group (AUC=0.860) (Z=2.274, 2.155; P=0.023, 0.031). In the validation set, the AUC of the deep learning model for diagnosing breast cancer was 0.892, which was comparable with that of the senior radiologist group (AUC=0.867) (Z=0.549, P=0.583); both were significantly higher than that of the junior radiologist group (AUC=0.750) (Z=2.429, 2.034; P=0.015, 0.042).

Conclusion The MRI-based deep learning model has good diagnostic value in breast cancer and can be used as a tool for clinical auxiliary diagnosis.