Detection of Suspicious Clusters in Women's Breast Image Using Convolutional Neural Network

Breast cancer is a type of cancer that develops from the breast tissue, and it is one of the leading causes of female deaths worldwide. Mammography is used to examine human breasts for screening and diagnosing the presence of breast cancer in the breast image. The recurrence of the use of mammography for screening produces enormous amounts of human breast data. Normally, radiologists interpret the screening mammograms, however, the process can be long and exhausting. As a result, the radiologist may not detect all breast cancers due to misinterpretation and unrevealing of hidden breast masses that can lead to cancer. Recent studies have investigated the current mammogram systems for effectively finding hidden breast masses, but none have given a conclusive solution. Hence, this research develops a deep learning framework that utilizes mean filtering inclined on Convolutional Neural Networks (CNN) to extract and train (learn) relevant features from breast images to detect anomalous breast masses. Experiments were conducted using a publicly available (Wisconsin) breast dataset, the result was compared with other popular existing methods indicating that new methods achieve superior performance when compared to other models with an Fl-score of 0.9761, a precision of 0.9881, and accuracy of98.64
^%
• The deployment of this model to the physician CAD can help in giving an accurate region in the breast where cancer is likely to occur in the future.