Multilevel Lacunarity Analysis Based Quantitative Parameter Sets for Gliomas Grade Classification
Year: 2026
Type: article
Abstract: In this study, a multilevel lacunarity analysis based quantitative parameter set is proposed for gliomas grade classification. The proposed feature set extracts multiscale texture information from segmented T2-weighted magnetic resonance images and characterize spatial heterogeneity of gliomas through lacunarity analysis. The lacunarity features are computed from multiple binary images generated using adaptive Otsu’s thresholding method. Lacunarity features at two spatial scales (r=2 and r=4) are computed from three threshold levels. A feature set of six multilevel features is constructed for glioma discrimination. Statistical analysis using student t-test demonstrated that multilevel lacunarity based texture features are significantly distinct between high grade (HG) and low grade (LG) glioma patients (p<0.05). The Cohen's d effect size values ranged from 1.252 to 1.706 confirming the strong discriminative capability of the proposed lacunarity-based features. Gliomas grade classification was accomplished with support vector machines (SVM), k-Nearest Neighbor (k-NN), linear discriminant analysis (LDA), Naive Bayes, subspace discriminant ensemble and decision tree classifiers. The results obtained shows that Naïve Bayes classifier perform better in comparison to other classifiers by providing accuracy of 92.0 ± 4.4 %, sensitivity of 90.0 ± 4.0 % and specificity of 95.0 ± 5.6 % for grade classification of Glioma patients. Our results suggest that the multilevel lacunarity based texture features identified in this study can discriminate LG and HG gliomas and can assist radiologists in classification of gliomas.
DOI: 10.54392/irjmt26517
Authors: Manu Gupta, Sravan K. Vittapu, Sakshi Gupta
Language: English
Subfield: Radiology, Nuclear Medicine and Imaging
Field: Medicine
Domain: Health Sciences
Open Access status: diamond
OpenAlex ID: W7214699412