Research map / Label-efficient and robust learning / Self-supervised and contrastive
Local Spatiotemporal Representation Learning for Longitudinally-consistent Neuroimage Analysis
Mengwei Ren, Neel Dey, Martin Andreas Styner, Kelly Botteron, Guido Gerig
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 |
|---|---|---|---|
| Use the proposed longitudinal pretraining and add segmentation consistency regularization | 3D U-Net · IBIS-wmgm one-shot | median IoU: 0.607 -> 0.676 | Improved |
| Use the proposed longitudinal pretraining without segmentation consistency regularization | 3D U-Net · OASIS3 one-shot | median HD95: 1.988 -> 1.940 | Improved |
| Add finetuning consistency loss (ablation) | proposed method · IBIS-wmgm one-shot | median IoU: 0.614 -> 0.676 | Improved |
| Add finetuning consistency loss (ablation) | proposed method · OASIS3 one-shot | median IoU: 0.739 -> 0.737 | Hurt |
| Increase projector and predictor MLP width from 256 to 2048 (ablation) | encoder-only similarity pretraining · IBIS-subcort | mean Dice: 0.829(0.068) -> 0.849(0.060) | Improved |
| Increase projector and predictor MLP width from 256 to 2048 (ablation) | encoder-only similarity pretraining · IBIS-wmgm | mean Dice: 0.733(0.062) -> 0.732(0.073) | Hurt |
| Add reconstruction loss to encoder-only similarity pretraining (ablation) | encoder-only similarity pretraining · IBIS-wmgm | mean Dice: 0.732(0.073) -> 0.713(0.079) | Hurt |
| Add decoder similarity layers, reconstruction loss, and augmentation (ablation) | proposed pretraining · IBIS-wmgm | mean Dice: 0.739(0.067) | Improved |
| Add orthogonality, variance, and covariance regularization with augmentation (ablation) | proposed pretraining · IBIS-subcort | mean Dice: 0.863(0.057) | Improved |
| Increase variance regularization μ from 10^-3 to 10^-2 (ablation) | regularized augmented pretraining · OASIS3 | mean Dice: 0.808(0.14) -> 0.813(0.13) | Improved |
| Add finetuning consistency loss (ablation) | proposed method · IBIS-subcort | mean Dice: 0.863(0.057) -> 0.870(0.052) | Improved |
| Add finetuning consistency loss (ablation) | proposed method · IBIS-wmgm | mean Dice: 0.758(0.062) -> 0.806(0.030) | Improved |
Closest papers by meaning
- Keypoint-Augmented Self-Supervised Learning for Medical Image Segmentation with Limited Annotation (NeurIPS 2023 · similarity 0.88)
- Autoregressive Sequence Modeling for 3D Medical Image Representation (AAAI 2025 · similarity 0.87)
- Beyond Instance-Level Self-Supervision in 3D Multi-Modal Medical Imaging (ICML 2026 · similarity 0.87)
- Integrating Sequence and Image Modeling in Irregular Medical Time Series Through Self-Supervised Learning (AAAI 2025 · similarity 0.85)
- Revisiting MAE Pre-training for 3D Medical Image Segmentation (CVPR 2025 · similarity 0.85)
- Adaptive Shrinkage Estimation for Personalized Deep Kernel Regression in Modeling Brain Trajectories (ICLR 2025 · similarity 0.85)
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.
