PED-SegNet: Modality-Aware Hybrid UNetR for Pediatric Brain-Tumor Segmentation on BraTS-PEDs

Background: Pediatric brain-tumor segmentation from multi-parametric MRI is challenging because of heterogeneous tumor characteristics, small or absent enhancing regions, and variations across imaging systems. Accurate delineation of the enhancing tumor (ET) region remains particularly difficult.Objectives: This study aims to develop PED-SegNet, a modality-aware deep learning framework for improving pediatric brain-tumor segmentation, boundary delineation, calibration, and robustness.Methods: PED-SegNet combines a transformer-based encoder with a UNet-style decoder to capture global context and fine spatial details. Squeeze-excitation gates and cross-modal attention enable adaptive fusion of MRI modalities. Class-balanced Dice loss, Focal Cross-Entropy, region-aware supervision for WT/TC/ET, and an ET absence-tolerant target were incorporated. The model was evaluated on the BraTS-PEDs ۲۰۲۳ dataset using subject-level five-fold cross-validation with Dice, HD۹۵, and ECE metrics, along with ablation and robustness analyses.Results: PED-SegNet achieved mean Dice scores of ۰.۹۰۸ (WT), ۰.۸۷۶ (TC), and ۰.۷۲۱ (ET), with HD۹۵ values of ۴.۲, ۵.۵, and ۷.۸ mm, respectively. ECE values were ۰.۰۳۱, ۰.۰۳۹, and ۰.۰۵۷. Compared with the baseline, Dice improved by ۰.۰۳۵, ۰.۰۴۱, and ۰.۰۷۸ for WT, TC, and ET, respectively, with a ۱.۲ mm reduction in HD۹۵. The model required approximately ۴۹.۹ M parameters, ۱۴۱ G MACs, ۲۶ s per subject, and ۷.۲ GB VRAM.Conclusions: PED-SegNet effectively combines transformer-based contextual learning, adaptive multi-modal fusion, and region-aware optimization for pediatric brain-tumor segmentation. The results demonstrate improved segmentation and boundary accuracy with practical computational requirements, indicating potential for deployment-oriented neuroimaging applications.

کلیدواژه ها:

Pediatric brain-tumor ، Transformer encoder ، UNet-style decoder ، Region-aware supervision ، Cross-modal attention