Towards Breast Cancer Response Prediction using Artificial Intelligence and Radiomics

With breast cancer being one of the recurring diseases affecting women around the globe, the World Health Organization disclosed that more than 620,000 women died from breast cancer in the world in 2018 alone, which represents approximately 15% of all female cancer deaths. Thus, breast cancer diagnosis presents one of the main challenges that need to get timely treatments. In this context, multiple image modalities, namely mammography, echography and magnetic resonance Imaging (MRI) are used for breast tumor diagnosis. One of the main treatments of this pathology is chemotherapy. However, several secondary effects (hair loss, osteoporosis, vomiting, etc.) can occur due this treatment, and cancer can not respond to it. This paper aims to suggest a novel method to predict breast tumor response to treatment, using three main steps: 1. Tumor segmentation from MR images ; 2. Extraction of features from segmented tumors in order to generate a complete and exploitable database ; 3. The use of deep and machine learning architectures to compute tumor-response prediction models. Experimental results are applied using a public QIN Breast DCE-MRI dataset of breast cancer patients.