Using Pectoral Muscle Removers in Mammographic Image Process to Improve Accuracy in Breast Cancer

Abstract

Cancer is a disease that attacks almost any organ or tissue of the body when abnormal cells grow uncontrollably and invade adjacent parts of the body. The second highest incidence of cancer in Indonesia is breast cancer with 42.1 cases per 100,000 population with an average mortality rate of 17 per 100,000 population. Mammography is a special imaging modality with x-rays to produce detailed breast images with at least 2 viewpoints, namely Craniocaudal (CC) or top view and Medio Lateral Oblique (MLO) or side view. The chest muscle area on the MLO display often interferes with the cancer identification process on mammography images because it has a dominant density and is similar to the density of cancer tissue. This research proposes a framework consisting of pectoral muscle detection on MLO display, image enhancement process, segmentation, and feature extraction. This study succeeded in increasing the accuracy of the MLO display mammography image after using the pectoral muscle remover using gradual edge detection and Hough lines Transform with the ratios of accuracy, precision, specificity, and sensitivity for images without pectoral muscle removers respectively were 33.59%, 30%, 11.49% and 80.48%. As for the images with pectoral muscle removers, the accuracy, precision, specificity, and sensitivity values respectively ​​were 68.67%, 64.71%, 57.14%, and 80.49%. For future projects, this research can be developed using Convolutional Neural Network (CNN) to improve accuracy. This is expected to help doctors and radiologists in the process of reading patient mammography so it can reduce mortality from breast cancer.

An Image Processing Framework for Breast Cancer Detection Using Multi-View Mammographic Images

“…There was a slight difference in the pre-processing stage between CC and MLO views. In the MLO view, the orientation of the image was determined and the pectoral muscles were removed, which is modified from the method according to [30]. In order to get the desired level of contrast, we first modified the adaptive histogram using Contrast Limited Adaptive Histogram Equalization (CLAHE), then we used the contrast stretching approach.…”