Work

Engineering Handcrafted Features for Machine-Learning-Based Segmentation of Brain Tumors in Multi-Spectral MRI Records

Year: 2026

Type: conference-paper

Abstract: To balance accuracy and efficiency in brain tumor segmentation, this study introduces an Markov Clustering-based feature selection strategy for ensemble learning. The method utilizes pairwise feature data from initial ensemble outputs to construct a graph, where Markov Clustering identifies the most effective feature subsets. Testing on BraTS 2019 low-grade tumor records demonstrates that the feature set can be reduced by 80% (down from 164 to 33 features) with negligible impact on accuracy. This approach significantly streamlines the segmentation process by eliminating redundant features while preserving the high discriminative power required for clinical applications.

Source: IEEE International Conference on Intelligent Engineering Systems (INES)

DOI: 10.1109/ines69513.2026.11661204

Authors: Ioan-Marius Pisak-Lukáts, Levente Kovács, László Szilágyi

Institution: Obuda University

Language: English

Topic: Brain Tumor Detection and Classification

Subfield: Computer Vision and Pattern Recognition

Field: Computer Science

Domain: Physical Sciences

SDG: Good health and well-being

Open Access status: hybrid

OpenAlex ID: W7204161282

FWCI:

Cites: 21

Cited by:

Related to: 10