Deep transfer learning radiomics based on two-dimensional ultrasound for predicting the efficacy of neoadjuvant chemotherapy in breast cancer

Purpose: We investigate the predictive value of a comprehensive model based on preoperative ultrasound radiomics, deep migration learning, and clinical features for pathological complete response (pCR) after neoadjuvant chemotherapy (NAC) for the breast cancer. Methods We enrolled 211 patients with pathologically confirmed the breast cancer who underwent NAC. The patients were randomly divided into the training set and the validation set in the ratio of 7:3. The deep learning and radiomics features of pre-treatment ultrasound images were extracted, and the random forest recursive elimination algorithm and the least absolute shrinkage and selection operator were used for feature screening and DL-Score and Rad-Score construction. According to multiple logistic regression, independent clinical predictors, DL-Score, and Rad-Score were selected to construct the comprehensive prediction model DLR + C. The performance of the model was evaluated in terms of its predictive effect, calibration ability, and clinical practicability. Result Compared to the clinical, radiomics (Rad-Score), and deep learning (DL-Score) models, the DLR + C accurately predicted the pCR status, with an area under the curve (AUC)of 0.906 (95% CI: 0.871–0.935) in the training set and 0.849 (95% CI: 0.799–0.887) in the validation set, with good calibration ability (Hosmer-Lemeshow: P > 0.05). Moreover, decision curve analysis confirmed that the DLR + C had the highest clinical value among all models. Conclusion The comprehensive model DLR + C based on ultrasound radiomics, deep transfer learning, and clinical features can effectively and accurately predict the pCR status of breast cancer after NAC, which is conducive to assisting clinical personalized diagnosis and treatment plan.