Research map / Architectures and efficiency / CNN design
FBTA: Enabling Single-GPU End-to-End Gigapixel WSI Classification with Feature Bridging and Translation Alignment
Jiuyang Dong, Jiahan Li, Junjun Jiang, Yongbing Zhang
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 FBTA to ABMIL | ABMIL · STAD (ResNet-50) | accuracy: +15.8% | Improved |
| add FBTA to MambaMIL | MambaMIL · NSCLC-Shot50 (ResNet-50) | accuracy: +4.1% | Improved |
| add FBTA to MIL with ViT-B/32 extractor | MIL models · STAD | AUC: +0.1% to +1.7% | Mixed |
| add FBTA to SimSiam-pretrained ResNet-50 ABMIL | ABMIL · NSCLC-Shot50 | AUC: +8.6% | Improved |
| add FBTA to SimSiam ResNet-50 ABMIL | ABMIL · NSCLC-Shot100 | F1: -0.2% | No clear effect |
| add FBTA to DTFD-AFS augmentation | DTFD-AFS ABMIL · NSCLC-Shot50 ABMIL | AUC: 74.4 -> 89.9 | Improved |
| use FBTA instead of Local Learning | Local Learning · NSCLC-Shot20 | AUC: 72.1 -> 82.1 | Improved |
| add frozen-feature bag to end-to-end training (ablation) | FBTA · NSCLC-Shot20 | average AUC: 73.2 -> 76.8 | Improved |
| add translated-feature bag (ablation) | FBTA · NSCLC-Shot20 | average AUC: 76.8 -> 77.6 | Improved |
| train with end-to-end bag only (ablation) | FBTA · NSCLC-Shot20 | average AUC: 74.3 -> 73.2 | No clear effect |
| make the extractor trainable (ablation) | FBTA · NSCLC-Shot20 ABMIL | AUC: 72.3 -> 77.0 | Improved |
| KMeans stratified instead of random pseudo-bag sampling (ablation) | FBTA · NSCLC-Shot20 ABMIL | AUC: 83.5 -> 83.4 | No clear effect |
Closest papers by meaning
- PathVQ: Reforming Computational Pathology Foundation Model for Whole Slide Image Analysis via Vector Quantization (NeurIPS 2025 · similarity 0.88)
- Single GPU Task Adaptation of Pathology Foundation Models for Whole Slide Image Analysis (NeurIPS 2025 · similarity 0.86)
- Fast and Accurate Gigapixel Pathological Image Classification with Hierarchical Distillation Multi-Instance Learning (CVPR 2025 · similarity 0.85)
- CPath-Omni: A Unified Multimodal Foundation Model for Patch and Whole Slide Image Analysis in Computational Pathology (CVPR 2025 · similarity 0.85)
- FOCUS: Knowledge-enhanced Adaptive Visual Compression for Few-shot Whole Slide Image Classification (CVPR 2025 · similarity 0.85)
- Generalizable Whole Slide Image Classification with Fine-Grained Visual-Semantic Interaction (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.
