Detection of lesions in breast image using median filtering and convolutional neural networks
Omobayo Ayokunle Esan, Munienge Mbodila, Patrick Mukeninay Madimba · 2024
Many people all over the world are impacted by cancer, a serious health issue. This illness has already claimed many lives and will do so in the future. Breast cancer has recently surpassed cervical cancer as the most prevalent cancer in women. At present, mammography screening is used for early detection of any form of lumps or lesions in breast images before developing cancer. Accurate screening and detection of breast lesions is a challenging issue for many medical practitioners even with the use of mammography imaging due to the process of interpretation of mammogram results, which are often done subjectively through visual analysis consequently leading to some breast lesions going unnoticed. The skewed median filtering technique on a convolutional neural network (CNN) is used to accurately and quickly identify breast lesions in their early stages. A publicly accessible breast dataset (Wisconsin) was used for the experiments. The results demonstrate that the proposed methods outperform other techniques in terms of F1 score, precision, recall, and accuracy with 0.9661, 0.9881, 0.9783, and 98.64%, respectively in comparison with the other methods. This model can help radiologists identify breast regions where cancer is most likely to develop in the future.