Detail Consistent Stage-Wise Distillation for Efficient 3D MRI Segmentation
For medical image segmentation practitioners, DCD enables efficient deployment of high-performing 3D segmenters by compressing models while retaining fine structural details, addressing a key bottleneck in memory- and latency-constrained settings.
The paper proposes Detail Consistent Distillation (DCD), a stage-wise distillation framework that preserves fine structural details in compressed 3D MRI segmentation models by aligning teacher-student features in the wavelet domain. Experiments on BraTS 2024 and ISLES 2022 benchmarks show superior segmentation performance without inference overhead.
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.