Design of Hybrid Adaptive Fractional Order Model with Multi-Criteria Objective Function for Enhanced Mammographic Image Segmentation in Breast Cancer Detection

Abstract

This paper presents a hybrid adaptive fractional-order DPSO framework for multi-threshold mammographic image segmentation that aims to enhance the delineation of dense tissue and abnormalities at their early stage. The method incorporates a multi-criteria objective function combining between-class variance, edge alignment and inter-channel coherence to enhance statistical separability and structural accuracy. It also utilizes adaptive scheduling of the inertia and acceleration coefficients for balancing exploration and exploitation during optimization and fractional-order velocity updates to introduce long-term memory for smoother and more stable convergence. The proposed framework is qualitatively and quantitatively evaluated using the Mini-MIAS dataset, considering improvements in segmentation accuracy, boundary preservation, convergence speed and consistency in comparison with Otsu’s thresholding, classical PSO and standard DPSO. Confirmatory experiments in both grayscale and perceptually uniform colour spaces reinforce the model’s capabilities in delivering coherent and clinically meaningful segmentations. The proposed approach lays a sound foundation for optimization-driven mammogram analysis and future computer-aided detection systems.

KEYWORDS

Hybrid adaptive fractional-order DPSO framework,multi-criteria objective function,Mammographic image Segmentation,Breast Cancer Detection,Dynamic Particle Swarm Optimisation.