Optimization of Breast Cancer Classification Using Faster R-CNN

Abstract

Breast cancer results from aberrant cell division in the breast and leads to the formation of tumors. The modern lifestyle, which is instant and rarely exercises, is the main driving force for this disease. Therefore, this study aims to diagnose by recognizing the specific characteristics of cancer in a benign or malignant class in the breast area. This study approach uses the deep learning technology model Faster R-CNN and dataset Mammographic Image Analysis Society (MIAS). This model requires unique image characteristics to recognize and produce a higher accuracy value. Furthermore, this study proposes optimizing an image segmentation approach using Matlab, ImageJ, and Python software to enrich cancer-specific images. This approach plays a vital role in increasing the accuracy of cancer detection. The results of this study before optimization have an accuracy rate of 63.47% using a smartphone camera; after optimization, the highest accuracy value becomes 90.43%, therefore 9.57% requires further examination by a specialist. Based on these results, these results help assist radiologists in making decisions about the results of the initial examination of breast mammogram data.