Hybrid Meta-Heuristic Optimized Multilayer Perceptron For Effective Breast Cancer Detection Based On The BIRAD Score Using Mammographic images

Abstract

Breast cancer remains one of the most prevalent malignancies affecting women globally. Mammography screening is one of the most effective methods for detecting early signs and symptoms of breast cancer. Timely diagnosis is crucial for effective treatment and to improve survival rates. Medical imaging techniques, particularly mammography, are commonly utilized for the detection of breast cancer and for monitoring tumour progression. However, this diagnostic technique often results in a relatively low positive predictive value for breast biopsies, leading to unnecessary procedures for abnormal findings. The Breast Imaging Reporting and Data System (BI-RADS) was established as a standardized tool for reporting mammographic results, categorizing unusual findings into specific groups. In this research ensemble models combined with optimization techniques to improve the detection process. Initially, pre-processing is carried out using a Gaussian Bilateral Filter (GBF) filter to eliminate noise and increase the mammography image quality. After that, the region of interest (ROI) is determined through Adaptive Histogram Thresholding and Contour Clustering (AHT-CC) algorithm. Next feature extraction is performed by applying Convolutional VGG-16 (ConV-16) model. Finally, the extracted feature is inputted to Hybrid Bayesian Genetic Optimized Multilayer Perceptron (HBGOMLP-EnML) for exact identification of breast cancer based on BI-RADS categories. The investigational analysis specifies that the proposed model attains accuracy of 98.6%, precision of 97.1%, recall of 97.2%, F1-score of 97.2% with different classifiers. Furthermore, the findings suggest that the proposed approach is competitive with other classification models documented in the literature, offering a practical and beneficial solution.