Breast Cancer Malignancy Prediction Using Incremental Combination of Multiple Recurrent Neural Networks

Abstract

Breast cancer is the most common cancer among women worldwide. An early detection of malignant of breast cancer, followed by proper treatment, can great improve the survival rate of patients. Recently, the deep learning based malignancy prediction models for breast cancer have been proposed. However, these models are usually trained with single type of clinical text, which are still not effective enough to predict breast cancer malignancy. In this paper, we follow the deep incremental learning framework and propose a prediction model of breast cancer malignancy by incremental combination of multiple recurrent neural networks. Specially, the model first uses multiple recurrent neural networks (RNNs) for generating features from the multi-types of clinical text including B-ultrasound, X-rays, Computed Tomography (CT), and Nuclear Magnetic Resonance Imaging (MRI), and then combines the generated features in an incremental way. Finally, we add one more recurrent neural network layer for classifying benign and malignant of breast cancer based on combined generated features.