Manuscript received March 3, 2025; revised March 27, 2025; accepted November 13, 2025; published July 16, 2026.

Abstract—Brain tumor detection remains a challenging problem due to the complexity of tumor structures and variability in Magnetic Resonance Imaging (MRI) data. Accurate detection and segmentation of brain tumors in medical imaging are essential for effective diagnosis and treatment planning. This study presents a multi-segmentation approach for brain tumor localization combining the Expectation-Maximization (EM) algorithm with an improved region-growing classification technique. To improve segmentation accuracy and computational efficiency, several quasi-random sampling methods were evaluated, including Halton, Sobol, Hammersley, Faure, and Poisson Disk Sampling. Experimental results demonstrate that the Sobol sampling method achieves superior precision in delineating tumor and edema regions. Its robustness and efficiency make it particularly suitable for complex medical imaging applications. Although other sampling techniques also produced promising results, Sobol consistently outperforms them in both speed and accuracy. In conclusion, the study highlights the importance of adapting segmentation strategies to the dataset characteristics, clinical requirements, and operator expertise to ensure optimal performance in real-world implementation.

Keywords—brain tumor segmentation, quasi-random sampling, sobol sequence, Magnetic Resonance Imaging (MRI), expectation-maximization, region-growing

Cite: Boucenna Sidahmed, Z. Chama, N. Taleb, B. Hachemi, and E. B. Bourennane, "Comparative Study of Quasi-random Sampling for Brain Tumor Segmentation in MRI Using Expectation-Maximization (EM) and Region Growing," Journal of Image and Graphics, Vol. 14, No. 4, pp. 567-583, 2026.

Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).