Predicting axillary lymph node metastasis in small breast cancers using convolutional neural networks for multiparametric magnetic resonance imaging (MRI).

AIM: This study aimed to develop a convolutional neural network (CNN) model using multiparametric magnetic resonance imaging (MRI) for accurate prediction of axillary lymph node metastasis (ALNM) in patients with small breast cancers.

MATERIALS AND

METHODS

Data of a total of 137 patients (ALNM: [n=47] and non-ALNM: [n=90]) with pathologically confirmed small breast cancers (<2 cm) from January 2018 to April 2022 were retrospectively analysed. Samples were randomly divided into training and validation sets in an 8:2 ratio. Diffusion-weighted imaging (DWI), T2-weighted image (T2WI), and dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) were cropped to obtain regions of interest for lesions. A CNN using the residual network (ResNet)-50 architecture was utilised for result prediction. Subsequently, the obtained results were fed into a selection process involving support vector machine (SVM) and random forest (RF) for integrated learning, resulting in the construction of an MRI ALNM multiparameter model. Accuracy, precision, recall, and receiver operating characteristic area under the curve (AUC) were used to evaluate model performance.

RESULTS

Regarding the validation cohort, the accuracy of the CNN prediction model for a single sequence was 76.50% to 81.50%; the AUC was 0.786 to .891. Considering the multiparameter ensemble model, DCE-MRI + T2WI using the RF classifier provided the best results. The AUCs of the training and test sets were 0.973 and 0.912, respectively.

CONCLUSION

The CNN model based on multiparameter MRI could be used to predict ALNM in small breast cancers. The RF model combining DCE-MRI and T2WI achieved better predictive performance, which may lead to more personalised treatment strategies before surgery for patients with small breast cancers and improve patient outcomes.