通过分阶段小波对齐，让轻量3D脑影像分割模型保持细节清晰。
Detail Consistent Stage-Wise Distillation for Efficient 3D MRI Segmentation
- 在小波域分阶段蒸馏方向性细节，保留细微结构
- 在BraTS 2024和ISLES 2022上优于现有轻量化方法
- 训练时使用，推理无额外开销，适合临床部署
高精度3D医学图像分割模型（如nnU-Net）的部署常受限于内存占用和推理延迟。压缩虽必要，但紧凑的3D编码器在多尺度下重复下采样易丢失细小病灶和锐利边界等结构细节。本文提出细节一致蒸馏（DCD），一种分阶段蒸馏框架，通过在小波分解表示中对齐教师-学生特征，保持跨尺度的结构细节。在每个编码器阶段，DCD仅蒸馏小波域的方向性细节成分，而粗略近似部分保持相对自由，避免全局语义过度正则化。DCD仅用于训练，不引入推理时开销。在BraTS 2024与ISLES 2022基准测试中，该方法在使用3D多模态数据的脑部MRI分割任务上表现优异。代码与实现细节已公开于https://github.com/ClinicaAlpha/DCD-3D-MedSeg。
原文摘要 · Abstract (English)
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.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。