2026· International journal of research and innovation in applied science· 0 citations
TL;DR
A formulation-driven segmentation framework is introduced that explic-itly integrates global structural modeling and boundary-aware constraints and yields a principled foundation for developing clinically reliable segmentation systems.
Abstract
Accurate brain tumor segmentation from multimodal magnetic resonance imag-ing (MRI) remains challenging, particularly in volume-level evaluation, where boundary leakage and volumetric fragmentation frequently occur. Although deep learning-based segmentation models achieve strong patch-level performance, their reliance on region-overlap optimization often leads to unstable global struc-ture and poor boundary coherence when applied to full volumetric data. In this work, a formulation-driven segmentation framework is introduced that explic-itly integrates global structural modeling and boundary-aware constraints. The proposed method introduces adaptive Fourier enhancement to stabilize volu-metric continuity by leveraging frequency-domain representations, followed by Dice-focused fine-tuning and multi-scale training to improve region filling and robustness to tumor size variability. Spatial gradient supervision is incorporated to enforce boundary geometry. Furthermore, a novel gradient–frequency coupling mechanism is designed to condition spectral modulation on boundary saliency, resolving the trade-off between boundary sharpness and volumetric smoothness. Extensive analyses on a benchmark dataset demonstrate that each component contributes progressively to the segmentation quality. The proposed formulation increases the volume-level whole-tumor Dice score from 0.0299 in the baseline configuration to 0.1772 in the final stage, representing a substantial improvement in volumetric consistency. Qualitative results further confirm that these gains arise from reduced fragmentation and improved global tumor structure rather than superficial pixel-level refinements. The results demonstrate the importance of formulation-level integration of geometric and spectral signs for robust volume-level brain tumor segmentation. They yield a principled foundation for developing clinically reliable segmentation systems.
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 and outperforms both classical active contour models and contemporary deep learning architectures in whole-tumor segmentation from...
M. Kazemi, H. Farsi, Aboozar Ghaffari et al.· Scientific Reports· 0 citations
Automated segmentation of post-treatment glioma from MRI is essential for accurate diagnosis and treatment planning, yet remains difficult. Surgery and chemoradiotherapy cause structural changes in brain like resection cavities, edema, and fibrosis, that alter imaging appearance and mimic tumor recurrence. Contemporary...
A new optimization-aware dual-domain framework for MIS (O2D-MedNet) is proposed, validating the effectiveness of the proposed O2D-MedNet for robust MIS under heterogeneous conditions and enhanced sensitivity, structural consistency, and optimization stability.
Shi-Peng Hu, Miao-Hui Wang, Jing Qin et al.· 0 citations
Accurate kidney and renal-tumor segmentation is challenging because lesion size, location, morphology, and boundary contrast vary substantially across abdominal CT scans. Most existing methods rely on a single form of local evidence and struggle to recover the boundaries of small lesions while maintaining global anatom...
Si-Yuan Liang, Cheng-Chuan Xu, Chao Lu et al.· Frontiers in Artificial Inte...· 0 citations
MGBR-Net is proposed, a morphology-prior guided structural boundary refinement network for polyp segmentation that achieves leading average mDice and mIoU performance and remains competitive across complementary region- and boundary-oriented metrics compared with 14 representative methods.
Shuai Zhang, Hao Fang, Xu-Yang Teng et al.· Journal of imaging informati...· 0 citations
CIRSeg combines 3D CutMix with stochastic intensity transfer using either Nyul augmentation or histogram matching to improve robustness to heterogeneous MRI intensities, and cascaded architecture decouples low-resolution anatomical localization from full-resolution boundary refinement.
Ru Xu, Ming-Qi Gao, Sheng-Da Luo et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.