- Home
- Archives
- 2026
- Vol. 7 Issue 6
- Article
An Enhanced Region Growing and Morphological Framework for MRI Brain Tumor Segmentation
Authors
1 2 3 Department of Computer Science and Engineering, Alakh Prakash Goyal University, Shimla, India.
Article Information
- Published In
- Volume 7, Issue 6 (2026)
- Published Online
- November-December 2026
- Pages
- 113-121
- Article Type
- Original Article
Abstract
View PDFAccurate segmentation of brain tumors from Magnetic Resonance Imaging (MRI) supports diagnosis, treatment planning and disease monitoring. Deep learning methods achieve high accuracy but depend on large annotated datasets and GPU-based training, which limits their use in low-resource settings, whereas classical region growing is lightweight and interpretable but is sensitive to noise and seed placement and often yields fragmented tumor masks. This paper presents a morphology-enhanced region-growing framework that combines Gaussian smoothing, center-of-mass seed initialization, intensity-based region growing and morphological post-processing (dilation with a disk-shaped structuring element and closing). The framework was implemented in MATLAB R2016b and evaluated on 50 publicly available grayscale MRI slices (512 × 512 pixels) from Kaggle, with the ground-truth masks supplied with the dataset, using the Dice Similarity Coefficient, Jaccard index, Precision, Sensitivity and Accuracy. Compared with baseline region growing, the proposed pipeline raised the mean Dice score from 0.0311 to 0.7015, the Jaccard index from 0.0201 to 0.5830, Precision from 0.1063 to 0.8215 and Sensitivity from 0.0207 to 0.7136, while Accuracy increased from 0.9668 to 0.9868. The proposed pipeline also outperformed thresholding, edge detection and watershed evaluated on the same images (mean Dice 0.0865, 0.0639 and 0.0374), and qualitative results show more compact, connected masks with fewer false-positive fragments. The method needs no training data or GPU and every stage can be inspected, making it a practical option for resource-constrained and educational settings; seed points derived from the ground truth and the small evaluation set are the main limitations and motivate automatic seeding and validation on larger multi-institutional datasets
Related Articles
Share Article
*Instagram doesn't support direct link sharing from web. Copy the link and share it in your Instagram story or post.