Enhanced Breast Cancer Detection Using Modified Particle Swarm Optimization-Based Back Propagation Neural Network

Abstract

Breast cancer detection faces challenges due to late diagnosis and inaccurate results, leading to improper treatment. Particle Swarm Optimization (PSO), a common optimization technique, often gets trapped in local optima, resulting in high false negatives and computational time. To address this, a study devised a modified PSO-based Back Propagation Neural Network (BPNN) using White Shark Optimization (WSO) for more accurate breast cancer detection. The research collected 1097 mammogram images from two medical centers and categorized them into benign, and malignant groups. Image preprocessing involved techniques like Otsu's method and contrast limited adaptive histogram equalization, with Fuzzy C Means used for segmentation. Region of Interest (ROI) extraction utilized Gabor filter and Local Binary Pattern, followed by weighted average fusion of textural features. WSPSO was then applied for optimal feature selection, and BPNN classified the identified features. MATLAB R2016a implemented the system. Evaluation metrics included accuracy, precision, sensitivity, specificity, False Positive Rate (FPR), and Computation Time (CT). Compared to PSO-BPNN, WSPSO-BPNN demonstrated superior performance. The accuracy, precision, sensitivity, specificity, FPR and CT of WSPSO-BPNN for malignant dataset were 96.88%, 99.08%, 96.43%, 97.92%, 2.08%, and 32.38 seconds, respectively, while the corresponding values for PSO-BPNN were 93.13%, 96.33%, 93.75%, 91.67%, 8.33% and 44.03 seconds, respectively. Also for the accuracy, precision, sensitivity, specificity, FPR and CT of WSPSO-BPNN for benign dataset were 96.15%, 98.14%, 96.34%, 95.71%, 4.29%, and 40.04 seconds, respectively, while the corresponding values for PSO-BPNN were 93.59%, 96.27%, 94.51%, 91.43%, 8.57% and 61.94 seconds respectively showcasing its potential for improving breast cancer diagnosis. Keywords: Back Propagation Neural Network, Breast Cancer, Classification, Mammogram Images, Particle Swarm Optimization, White Shark Optimization. Oyerinde, I.M., Olabiyisi, S.O., Falohun, A.A. & Awodoye, O.O. (2024); Enhanced Breast Cancer Detection Using Modified Particle Swarm Optimization-Based Back Propagation Neural Network. Journal of Advances in Mathematical & Computational Science. Vol. 12, No. 1. Pp 53-62. Available online at www.isteams.net/mathematics-computationaljournal. dx.doi.org/10.22624/AIMS/MATHS/V12N1P5