Multi-Attention Convolution Neural Network to Detect Breast Cancer

Abstract

Medical Imaging is one of the most effective approaches to diagnose the disease than biomarker examination or biopsy specimen study. Mammogram images are low intense digital imaging and Communication in Medicine Standards. Advanced image processing techniques are used to identify the abnormalities on the breast by applying machine learning algorithms to train the model and advance deep learning concepts are used to identify and classify tumors and class of tumors. Enhanced images will improve the detection rate of tumors, The proposed methodology used to enhance a mammogram image is TopHat Transformation and Segmentation used to detect the boundaries and feature extractions method using Wavelet-based clustering. The Tumours are classified using MACNN classification use to classify BENGIN and MALIGNANT. The proposed model obtained 93.36% of precision,93.30% Recall rate, and 93.22% gained accuracy. The Accuracy Obtained with the proposed method gains 93.30% and 93.22%. The same dataset with other approaches obtained less than 80% accuracy. AlexNet obtained 74.24%, VGG16 obtained 67.42%, GoogleNet obtained 68.94%, ResNet18 obtained 78.03%, and InceptionResNet V2 obtained 68.18% of accuracy.