Abstract

One important prognostic marker in breast cancer that affects treatment choices and patient outcomes is lymph node metastasis (LNM). Promising opportunities for automating and improving LNM prediction using medical data are presented by recent developments in deep learning. In order to predict lymph node metastases in patients with breast cancer, this research suggests a multimodal method that combines structured clinical data with imaging modalities including mammography, ultrasound, and magnetic resonance imaging (MRI). We create and assess a deep learning framework that combines fully connected networks processing clinical information with convolutional neural networks (CNNs) for picture analysis. Our model outperforms conventional machine learning baselines in accuracy, sensitivity, and area under the ROC curve (AUC), demonstrating superior predictive performance. The multimodal approach shows how deep learning may help with individualized diagnosis and treatment planning by accurately and noninvasively predicting LNM status.