Honey badger algorithm-optimized morphological segmentation combined with Gabor–KNN classification for meningioma brain tumor

This study presents a novel approach for the segmentation and classification of small meningioma brain tumors located near the perichondrium, a region where conventional methods often fail due to subtle intensity variations and complex tissue boundaries. The proposed methodology integrates Gabor filtering for enhanced feature extraction with a K-Nearest Neighbors (KNN) classifier, optimized by the honey badger algorithm to refine morphological segmentation parameters and distance metrics. By combining multi-scale, multi-orientation Gabor filters with probability-based morphological operations, the approach effectively captures textural heterogeneity while suppressing background noise. Experimental results demonstrate significant improvements over traditional methods, achieving near-perfect segmentation agreement with expert contours, with a DICE coefficient of 93.7% at k = 7—a 7.4% improvement—and superior precision (83.21%), recall (84.5%), and AUC (84.2%). The method also exhibits reduced computational complexity, robustness with limited training data, and perfect classification (F1 = 1) for highly confident predictions. The framework delivers greatest clinical value in challenging anatomical locations where intensity-based methods prove inadequate. In conclusion, the honey badger algorithm-optimized morphological segmentation combined with Gabor–KNN classification offers a clinically valuable tool for computer-aided diagnosis, with potential for extension to other tumor types and integration with deep learning architectures.