Use Case of ML-Algorithm to Effectively Detect Malignant Breast Tumors Without Compromising Medical Data Privacy

Magnetic Resonance Imaging (MRI) scans are used to show suspicious lesions in patients with breast tumors. The treating physicians extract information from the imaging about the damage that occurs in each patient and after doing a biopsy on the tumor, they decide on the type of cancer. The purpose of this paper is to showcase the development of prediction algorithms using machine learning to predict histological markers for patients with breast cancer without performing the biopsy upon further development of these markers. Taking parts of the information extracted from the MRI scans, various attempts were made to predict these indicators as accurately as possible. Five algorithms were written in MATLAB using linear interpolation. Before the execution of the linear interpolation the data were sorted in the most efficient way possible so that the algorithm was trained with data of the greatest possible variety of patient cases. After the execution of the algorithm, a 91.67% accuracy was achieved on the Tumor Grade marker and 100% accuracy on the ER, PR and CERB-2 markers.