Quality improvement and segmentation of mammography images using Python

Mammography is one of the early examinations to detect the presence of breast cancer. In this research, mammography image quality was improved by reducing the degree of gray and noise and segmenting the image to clarify the cancer area. A total of seven digital mammography images were used with different types of tissue classifications derived from the database on the Kaggle.com website and processing was carried out using the Python programming language. The first step is to improve the quality of mammography images with the Anisotropic Diffusion method. Then testing the results of image quality improvement will be carried out by calculating the value of Mean Square Error (MSE) and Peak Signal to Noise Ratio (PSNR). In addition to improving image quality, this research also carried out image segmentation, image segmentation process using 2 methods are Otsu and multilevel Otsu. Meanwhile, testing the results of image segmentation will be carried out by calculating the Miclassified Area Mutual Overlap (MMO) value from manual segmentation and automatic segmentation. The test results of image quality improvement obtained good values where the MSE is getting closer to 0 and the PSNR value has reached 30 dB using a kappa value of 40 and a gamma of 0.005. The results of segmentation testing using the Otsu method, the images that pass the segmentation test are images with classifications of circumscribed, architectural, spiculated, and ill-defined masses. The results of segmentation testing using Otsu's multilevel method on asymmetry, calcification and ill-defined masses of image classification in each class variant have passed the segmentation test.