Research map / Label-efficient and robust learning / Self-supervised and contrastive
MOOZY: A Patient-First Foundation Model for Computational Pathology
Yousef Hassan, Vincent Quoc-Huy Trinh, Christopher Pal, Mahdi S. Hosseini
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 case aggregator (case transformer) to MOOZY (ablation) | MOOZY · Sixteen held-out tasks, macro | weighted F1: 0.749 -> 0.769 | Improved |
| add case aggregator (case transformer) to MOOZY (ablation) | MOOZY · Sixteen held-out tasks, macro | weighted ROC-AUC: 0.737 -> 0.763 | Improved |
| add case aggregator (case transformer) to MOOZY (ablation) | MOOZY · Sixteen held-out tasks, macro | balanced accuracy: 0.682 -> 0.702 | Improved |
| add Stage 1 SSL pretraining to Stage 2 supervision (no case aggregator) (ablation) | Stage 2 slide encoder · Sixteen held-out tasks, macro | weighted ROC-AUC: 0.721 -> 0.737 | Improved |
| add Stage 1 SSL pretraining to Stage 2 supervision (no case aggregator) (ablation) | Stage 2 slide encoder · Sixteen held-out tasks, macro | weighted F1: 0.743 -> 0.749 | Improved |
| add case aggregator when trained from scratch without SSL (ablation) | Stage 2 only · Sixteen held-out tasks, macro | weighted F1: 0.743 -> 0.731 | Hurt |
| add Stage 2 multi-task supervision to Stage 1 SSL encoder (no case aggregator) (ablation) | Stage 1 only · Sixteen held-out tasks, macro | weighted ROC-AUC: 0.715 -> 0.737 | Improved |
| use MOOZY instead of CONCH v1.5 MIL baseline | MOOZY · Sixteen held-out tasks, macro | weighted F1: 0.740 -> 0.769 | Improved |
| use MOOZY instead of PRISM | MOOZY · Sixteen held-out tasks, macro | weighted F1, balanced accuracy, weighted ROC-AUC: +4.19%, +7.93%, +6.95% | Improved |
| use MOOZY instead of TITAN | MOOZY · Sixteen held-out tasks, macro | weighted ROC-AUC: 0.768 -> 0.763 | Hurt |
| use MOOZY instead of TITAN on molecular subtype tasks | MOOZY · BRCA PAM50 subtype (TCGA-BRCA) | weighted F1: 0.72 -> 0.63 | Hurt |
| use MOOZY instead of best baseline on Residual Cancer Burden | MOOZY · Residual Cancer Burden (BC Therapy) | weighted F1: 0.51 -> 0.56 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- TITAN — TITAN: Query-Token based Domain Adaptive Adversarial Learning (ICCV 2025) · better on 1 of 5 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- PathVQ: Reforming Computational Pathology Foundation Model for Whole Slide Image Analysis via Vector Quantization (NeurIPS 2025 · similarity 0.88)
- Morphological Prototyping for Unsupervised Slide Representation Learning in Computational Pathology (CVPR 2024 · similarity 0.87)
- FOCUS: Knowledge-enhanced Adaptive Visual Compression for Few-shot Whole Slide Image Classification (CVPR 2025 · similarity 0.86)
- Unsupervised Foundation Model-Agnostic Slide-Level Representation Learning (CVPR 2025 · similarity 0.86)
- SlideChat: A Large Vision-Language Assistant for Whole-Slide Pathology Image Understanding (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)
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.
