Segmentation of brain tumors using a morphological Laplacian-enhanced active contour model on MRI

Abstract

Accurate segmentation of brain tumors in magnetic resonance imaging (MRI) is crucial for effective treatment planning and post-treatment monitoring. However, challenges such as intensity inhomogeneity, low-contrast boundaries, and highly irregular tumor morphology significantly complicate this task. In this work, we introduce a two-stage hybrid framework that integrates a convolutional U-Net for coarse tumor localization with a morphological Laplacian-enhanced active contour model for precise boundary refinement. While the U-Net is trained using bimodal input (FLAIR + T1ce) to generate reliable initial masks, all refinement steps and final performance evaluations are conducted exclusively on single-modality FLAIR images—ensuring clinical simplicity and robustness. The proposed active contour incorporates a novel local fitting term based on morphological Laplacian operators, which capture structural transitions and enhance edge sensitivity. An adaptive weighting function dynamically balances the influence of local and global intensity information during contour evolution, improving resilience to noise and intensity variations without compromising computational efficiency. When evaluated on the BraTS 2020 dataset, the proposed method achieves a Dice coefficient of 96.48% ± 1.57%, a Jaccard index of 93.43% ± 2.92%, and a sensitivity of 97.01% ± 1.69%. Comparative experiments demonstrate that the proposed approach outperforms both classical active contour models and contemporary deep learning architectures in whole-tumor segmentation from FLAIR-only MRI. The source code will be made publicly available upon acceptance.