Improving Breast Cancer Detection in Mammogram Images Using Moth Flame Algorithm Optimized Convolutional Neural Network
M. Sarathkumar, K. S. Dhanalakshmi, P. Richard Rajkumar · 2023
A woman's milk duct lining cells that have abnormal discontinuities can indicate breast cancer. Large numbers of breast cancer victims have symptoms in the milk ducts before passing away from the disease. The fatality rates could be lowered if the discovery is done quickly. Medical professionals and radiologists typically find it more challenging to identify breast cancer in mammography pictures. This research offers a Convolutional Neural Network based on Moth Flame Optimisation (CNN-MFO) for identifying breast cancer from mammography pictures in order to minimise manual assessment and streamline the job of classification. Because of the low quality of the acquired mammogram image, preprocessing is the most significant stage in mammography evaluation. The adaptive median filter have been developed in this work for finding a compromise between all-pass filtering and simple averaging with the goal to minimise the blurring effect of image. The adaptive median filter outperforms in terms of speckle reduction while keeping the edges. Furthermore, for separating the breast region from mammography pictures, the U-Net model is employed. This strategy assists radiologists in early detection and improves the effectiveness of the proposed system. Following segmentation, the optimal second order statistical features are extracted by employing GLCM method. Finally, the proposed MF optimized CNN classifier accurately predict the breast cancer from the mammogram images. Thus, the obtained results prove that, the proposed system outperforms than other existing methods.