Soft Clustering Technique for Brain Tumor Segmentation within Neutrosophic Framework

Abstract

Precise segmentation of brain tumors is vital in healthcare because it impacts diagnosis, treatment
planning, and patient outcomes. However, the brain has complex structures with non-linear and inhomoge
neous forms, which poses substantial challenges for conventional segmentation approaches. Despite of intensive
research works, there persists a need to improve segmentation accuracy while concurrently lowering the com
puting costs. To address this, this study presents an innovative segmentation framework that incorporates
kernel distance into the neutrosophic c-means (NCM) algorithm, together with the Kullback-Leibler (KL) di
vergence measure. Furthermore, the integration of kernel distance enables the algorithm to adeptly identify
non-linear structural fluctuations and eliminate outliers, which improves robustness. Besides, the KL divergence
mechanism accelerates convergence and improves segmentation precision by increasing the clustering process by
reducing the distance between neighbourhood membership degrees. The proposed approach has been assessed
against five established clustering algorithms utilizing both objective performance metrics and subjective visual
evaluations. Experimental findings validate that the proposed method attains enhanced accuracy with increased
computational efficiency and more precise delineation of tumor areas. Additionally, these results underscore
the potential of the developed method for brain tumor segmentation, helping to bridge existing methodological
gaps and promote more efficient medical image analysis

DOI 10.5281/zenodo.18675653