Enhanced Crow Search Optimization Deep Convolution Network Classifier for Early Detection of Breast Cancer Detection

Abstract

The worldwide women mainly affecting cancers diagnosed in Breast cancer through important implications for communal health. Advances in health check treatments, along with premature detection, have led to improved survival outcomes for patients. The early detection of breast cancer is primarily achieved by digital mammography, which is current standard viewing method. The conventional deep learning models is not perfect, leading to inaccuracies result and even unnecessary treatments for patients, or missed diagnoses that delay critical treatment. In order to overcome this challenge, a novel Enhanced Crow Search Optimization Deep Convolution Network Classifier (ECSODCNC) model is developed to aim of improve the breast cancer detection accuracy within minimal time as well as error rate in early stage. The proposed method comprises the five different procedures and it integrated into proposed DCNN classifier provide to accurate disease prediction. Proposed ECSODCNC utilizes DCNN which comprises of numerous layers namely input, convolutional layer, maxpooling and fully connected layer and output layer. These mammogram images are specified to input layer of the deep learning method. Next, the hidden layers are utilized to accurate computation by using mammogram images. After that, image preprocessing is designed in convolutional layer to improve the image quality using the Hamann indexive variation denoising method. The quality-enhanced mammogram images are then transferred to the max-pooling level to minimize dimensionality of the aspect maps when retaining mainly important data. In this layer, ROI segmentation and geometric features are removed and ROI are transmitting to the fully connected layer with minimum time consumption. In classification process, Canonical Concordance Correlation is performed. Enhanced Crow Search Optimization (ECSO) is also employed for the fine-tuning process of the CNN to optimize hyper parameters to improve model disease prediction performance and minimize error. Finally, the result of breast cancer classification is achieved at output layer. The performance of the proposed ECSODCNC model was experimentally analyzed based on various performance metrics. Quantitatively analyzed outcomes describe to performance of ECSODCNC model achieves the improved breast cancer detection accuracy with lesser time consumption than the different existing methods.