Research map / Label-efficient and robust learning / Self-supervised and contrastive
Big Self-Supervised Models Advance Medical Image Classification
Shekoofeh Azizi, Basil Mustafa, Fiona Ryan, Zachary Beaver, Jan Freyberg, Jonathan Deaton, Aaron Loh, Alan Karthikesalingam, Simon Kornblith, Ting Chen, Vivek Natarajan, Mohammad Norouzi
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 in-domain unlabeled Derm pretraining on top of ImageNet pretraining (ablation) | SimCLR ResNet-152 (2x) · Dermatology | top-1 accuracy (%): 66.38 -> 68.30 | Improved |
| add in-domain unlabeled CheXpert pretraining on top of ImageNet pretraining (ablation) | SimCLR ResNet-152 (2x) · CheXpert | mean AUC: 0.7671 -> 0.7729 | Improved |
| add MICLe multi-image positive pairs (ImageNet->Derm) (ablation) | SimCLR ResNet-50 (4x) · Dermatology | top-1 accuracy (%): 67.63 -> 68.81 | Improved |
| add MICLe multi-image positive pairs (Derm only) (ablation) | SimCLR ResNet-50 (4x) · Dermatology | top-1 accuracy (%): 66.93 -> 67.55 | Improved |
| add MICLe multi-image positive pairs (ImageNet->Derm) (ablation) | SimCLR ResNet-152 (2x) · Dermatology | top-1 accuracy (%): 68.30 -> 68.43 | Improved |
| use partial instead of full augmentation in MICLe (Derm) (ablation) | MICLe ResNet-152 (2x) · Dermatology | top-1 accuracy: 0.6697 -> 0.6761 | Improved |
| use partial instead of full augmentation in MICLe (ImageNet->Derm) (ablation) | MICLe ResNet-152 (2x) · Dermatology | top-1 accuracy: 0.6928 -> 0.6889 | Hurt |
| replace supervised ImageNet pretraining with SimCLR ImageNet->Derm | ResNet-152 (2x) · Dermatology | top-1 accuracy (%): 63.36 -> 69.43 | Improved |
| replace BiT ImageNet-21k supervised pretraining with MICLe ImageNet->Derm | ResNet-152 (2x) vs ResNet-101 (3x) · Dermatology | top-1 accuracy (%): 68.45 -> 70.02 | Improved |
| replace BiT ImageNet-21k supervised pretraining with SimCLR ImageNet->CheXpert | ResNet-152 (2x) vs ResNet-101 (3x) · CheXpert | mean AUC: 0.7720 -> 0.7729 | No clear effect |
| scale up the architecture from ResNet-50 (1x) to ResNet-152 (2x) with ImageNet->Derm pretraining | SimCLR · Dermatology | top-1 accuracy (%): 63.44 -> 68.30 | Improved |
| add in-domain CheXpert pretraining on top of ImageNet for ResNet-50 (1x) (ablation) | SimCLR · CheXpert | mean AUC: 0.7630 -> 0.7670 | No clear effect |
Closest papers by meaning
- Systematic Comparison of Semi-supervised and Self-supervised Learning for Medical Image Classification (CVPR 2024 · similarity 0.87)
- Contrastive learning of global and local features for medical image segmentation with limited annotations (NeurIPS 2020 · similarity 0.87)
- ACPL: Anti-Curriculum Pseudo-Labelling for Semi-Supervised Medical Image Classification (CVPR 2022 · similarity 0.87)
- Novel Class Discovery in Chest X-rays via Paired Images and Text (AAAI 2024 · similarity 0.86)
- A Unified Visual Information Preservation Framework for Self-supervised Pre-training in Medical Image Analysis (TPAMI 2023 · similarity 0.86)
- Continual Self-supervised Learning: Towards Universal Multi-modal Medical Data Representation Learning (CVPR 2024 · 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.
