Abstract:
Breast cancer, a common ailment among women worldwide, is commonly discovered by the emergence of a lump in the breast. The most severe and devastating type of cancer; it is produced by the normal proliferation of cells in mammary glands. In this research work, Automated Breast Cancer Diagnosis Optimized with Higher-Order Attribute-Enhancing Heterogeneous Graph Neural Networks Using Mammogram Images (ABCD-HAHGNN-MI) is proposed. The input images are collected from DDSM and CBIS-DDSM dataset. The input images are pre-processed using Distributed Nonlinear Polynomial Graph Filter(DNPGF) is usedto enhance their quality and remove noise from mammogram images. The segmented ROI is given into Quaternion Offset Linear Canonical Transform (QOLCT) for extracting features like Statistical, texture features, shape features, GLCM features.The extracted features are given into HAHGNNfor Automatic Breast Cancer Diagnosis, which classifies breast cancer as normal, benign, or malignant. Generally, HAHGNN doesn't express some adaption of optimization strategies for determining optimal parameters to assure accurate detection. Consequently, Bitterling Fish Optimization (BFO) is proposed to enhanceweight parameters of HAHGNN. The proposed technique isactivated in Python under some performance metrics, like accuracy, specificity, sensitivity, FI-score, ROC. The simulation outcomes display that the performance of the ABCD-HAHGNN-MI approach attains 29.8 percent, 21.2 percent, and 18.9 percent higher Accuracy, 24.7 percent, 32.5 percent, and 29.6 percent higher f-score, 25.8 percent, 28.5 percent, and 21.6 percent higher Sensitivity are analysed with existing methods likedeep learning depend capsule neural network method for breast cancer diagnosis using mammogram images (BCD-CNN-MI), Breast Cancer Detection Utilizing Deep Learning(BCD-DL),estimation of deep learning methods for identifying breast cancer utilizing is to pathological mammograms images(BCD-DL-HMI) respectively.
