D3Seg: Dependency-Aware Diffusion for Brain Tumor Segmentation with Missing Modalities

A new brain tumor segmentation model called D3Seg maintains stable performance even when standard MRI sequences are unavailable, according to research posted on the arXiv preprint server [1]. The model addresses a common clinical problem: incomplete multi-parametric MRI acquisition degrades the accuracy of existing segmentation methods [1]. The work, submitted by Danish Ali and revised on June 23, 2026, introduces three technical components designed to compensate for missing modalities [1]. A Multi-hop Modality Graph Fusion (MMGF) module models higher-order inter-modality dependencies, while a lightweight diffusion-based imputation mechanism compensates for missing T1ce and FLAIR feature representations in latent space [1]. A probability-space decision refinement step mitigates dominant-class overconfidence and improves delineation of underrepresented tumor subregions [1]. The researchers evaluated D3Seg on the BraTS 2023 Glioma dataset as a primary benchmark and further tested it on a subset of the external BraTS 2023 Meningioma cohort to assess generalization across tumor pathologies [1]. Compared with the current state-of-the-art model, D3Seg achieved approximately 1.5-2.0% Dice improvement on enhancing tumor (ET) and around 1.0% on tumor core (TC) across multiple missing-modality configurations on the glioma dataset [1]. Cross-cohort evaluation on the meningioma dataset showed consistent improvements in the challenging TC and ET regions, with approximately 1.5-3.0% and 1.5-6.5% gains respectively across several missing-modality configurations [1]. The paper appears on arXiv, an open-access repository of electronic preprints that, as of November 2024, receives about 24,000 submissions per month and hosts more than two million articles [6]. arXiv content is moderated but not peer-reviewed before posting [6]. The repository, founded in 1991, serves fields including computer science, mathematics, and physics, and has become a primary distribution channel for machine-learning research [6]. The D3Seg manuscript was posted in the Computer Vision and Pattern Recognition category [1].