Research map / Detection and segmentation / CNN features
Rethinking Model Calibration through Spectral Entropy Regularization in Medical Image Segmentation
Kun Cheng, Yukun Zhang, William Henry Nailon, Tonggang Zhao
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add temporal smoothing of the spectral profile to spectral entropy regularization (ablation) | LCE w/ LSpectral · BraTS2020 | ECE: 0.0065 -> 0.0015 | Improved |
| add temporal smoothing of the spectral profile to spectral entropy regularization (ablation) | LCE w/ LSpectral · FLARE2021 | ECE: 0.0170 -> 0.0008 | Improved |
| add temporal smoothing of the spectral profile (ablation) | LCE w/ LSpectral · FLARE2021 | HD95: 6.15 -> 4.55 | Improved |
| add temporal smoothing of the spectral profile (ablation) | LCE w/ LSpectral · BraTS2020 | DSC: 0.870 -> 0.872 | Improved |
| add unsmoothed spectral entropy regularization to CE (ablation) | U-Net CE baseline · BraTS2020 | ECE: 0.0091 -> 0.0065 | Improved |
| add unsmoothed spectral entropy regularization to CE (ablation) | U-Net CE baseline · FLARE2021 | ECE: 0.0255 -> 0.0170 | Improved |
| use spectral entropy regularization instead of CE alone | U-Net · ACDC | ECE (x10^-3): 32.5 -> 2.1 | Improved |
| replace CE with Focal Loss | U-Net · FLARE2021 | DSC (%): 91.5 -> 87.7 | Hurt |
| increase regularization weight lambda to 0.100 (ablation) | LCE w/ LSpectral · BraTS2020 | DSC: 0.872 -> 0.865 | Hurt |
| reduce lambda to 0.001 (ablation) | LCE w/ LSpectral · BraTS2020 | ECE: 0.0015 -> 0.0045 | Hurt |
| enlarge smoothing window W to 100 (ablation) | LCE w/ LSpectral · BraTS2020 | ECE: 0.0015 -> 0.0022 | Hurt |
| shrink smoothing window W to 10 (ablation) | LCE w/ LSpectral · BraTS2020 | ECE: 0.0015 -> 0.0025 | Hurt |
Closest papers by meaning
- Energy-guided Dual Domain-invariant Prompting Framework with Fourier Regularization for Generalized Few-Shot Medical Segmentation (AAAI 2026 · similarity 0.86)
- Harmonized Feature Conditioning and Frequency-Prompt Personalization for Multi-Rater Medical Segmentation (CVPR 2026 · similarity 0.85)
- S2M-Net: Spectral-Spatial Mixing with Morphology-Aware Adaptive Loss for Medical Image Segmentation (ICML 2026 · similarity 0.85)
- beta-FFT: Nonlinear Interpolation and Differentiated Training Strategies for Semi-Supervised Medical Image Segmentation (CVPR 2025 · similarity 0.83)
- DyCON: Dynamic Uncertainty-aware Consistency and Contrastive Learning for Semi-supervised Medical Image Segmentation (CVPR 2025 · similarity 0.83)
- S2S2: Semantic Stacking for Robust Semantic Segmentation in Medical Imaging (AAAI 2025 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
