Detail Consistent Stage-Wise Distillation for Efficient 3D MRI Segmentation
Quick Take
The Detail Consistent Distillation (DCD) framework enhances 3D MRI segmentation by preserving structural details across scales, achieving superior results on BraTS 2024 and ISLES 2022 benchmarks without increasing inference time. This method effectively distills directional detail components while maintaining coarse approximations, addressing the limitations of existing compact 3D encoders like nnU-Net.
Key Points
- DCD aligns teacher-student features in a wavelet-decomposed representation.
- The method distills directional detail components at each encoder stage.
- DCD introduces no overhead during inference, optimizing performance.
- Experiments show superior MRI segmentation results on BraTS 2024 and ISLES 2022.
- Code and implementation details are publicly available on GitHub.
Article Excerpt
From source RSS / original summaryarXiv:2605. 26382v1 Announce Type: new Abstract: Deploying high-performing 3D medical image segmenters (e. g. , nnU-Net) is often limited by memory footprint and inference latency. Compression is therefore necessary, but compact 3D encoders tend to lose fine structural cues (small lesions and sharp boundaries) as downsampling repeats across multi-resolution stages.
We propose Detail Consistent Distillation (DCD), a stage-wise distillation framework that preserves structural detail across scales by aligning teacher-student features in a wavelet-decomposed representation. At each encoder stage, DCD distills directional detail components in the wavelet domain while leaving the coarse approximation comparatively unconstrained, avoiding over-regularization of global semantics. DCD is used only during training and introduces no inference-time overhead.
Experiments on the BraTS 2024 and ISLES 2022 benchmarks demonstrate that our approach achieves superior performance in MRI segmentation using 3D multi-modal data. Code and implementation details for DCD are publicly available at https://github. com/ClinicaAlpha/DCD-3D-MedSeg.
Reader Mode unavailable (could not extract clean content).
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CV
See more →Evi-Steer: Learning to Steer Biomedical Vision-Language Models through Efficient and Generalizable Evidential Tuning
Evi-Steer introduces a novel evidential tuning framework for BiomedCLIP, achieving 0.11% parameter updates while enhancing uncertainty-aware fine-tuning. It outperforms state-of-the-art methods across 15 biomedical imaging datasets, proving effective in few-shot learning and domain shifts for clinical applications.