Driving · 2026-11-09
ProtoCouncil: Multi-Encoder Anatomical Prototypes for Prostate Cancer Detection in Micro-Ultrasound
Mohammad Mahdi Abootorabi, Nooshin Maghsoodi, Armin Saadat, Obed Dzikunu, Zhuoxin Guo, Hannes Cash, Ferdinand Luger, Adam Kinnaird, Brian Wodlinger, Parvin Mousavi, Purang Abolmaesumi
AI-assisted summary based on the paper. Consult the linked source for methods and limitations. Benchmark numbers appear after editor verification.
TL;DR
ProtoCouncil uses three foundation encoders (MedSAM, DINOv3, V-JEPA 2.1) to learn anatomical, case-based class prototypes at whole-gland, needle-track, and gland-sector levels for micro-ultrasound prostate cancer detection. It projects prototypes onto real training regions, fuses per-encoder class logits additively, and abstains when encoder predictions disagree.
Why it matters
The authors target traceability and uncertainty in core-level prostate cancer detection by grounding predictions in similarity to identifiable training prototypes with anatomical indexing. Encoder disagreement is used as a selective prediction signal to flag unreliable decisions and improve performance on retained cores.
Method
- Three frozen foundation encoders (MedSAM, DINOv3, V-JEPA 2.1) learn case-based prototypes at three anatomical levels: whole gland, needle track, and six gland sectors.
- Prototypes are projected onto real training-core regions; fusion keeps each class logit as a bias plus an exact sum of prototype similarity contributions from each encoder.
- Selective prediction: the model returns the fused prediction only when all three encoders agree; otherwise it abstains.
Limitation
The paper reports that cohort C includes only six patients with csPCa.
Abstract (from arXiv)
Purpose: Micro-ultrasound ($\mu$US) is a promising modality for prostate cancer (PCa) detection, but identifying suspicious tissue depends heavily on clinical experience. Models trained on sparse, noisy core-level pathology offer little insight into their decisions and can fail under domain shift. Clinical use needs a traceable decision process and a flag for uncertain predictions. Methods: We introduce ProtoCouncil, in which three foundation encoders (MedSAM, DINOv3 and V-JEPA 2.1) learn case-based prototypes at three anatomical levels: whole gland, needle track and gland sectors. Prototypes are projected onto real training-core regions, and label-free fusion keeps every class logit as a bias plus an exact sum of prototype similarity contributions. Encoder disagreement triggers selective prediction. Results: In five-fold patient-level cross-validation (6,588 cores, 693 patients), ProtoCouncil reaches 82.9% AUROC for clinically significant PCa (csPCa, Grade Group $\geq$3), 4.3 points above the best previously published method. In zero-shot evaluation on two external cohorts, ProtoCouncil achieves higher all-cancer and csPCa AUROC than all previous methods. Nearest same-class exemplars match test-core tumor involvement better than random same-class cases. Externally, csPCa accuracy falls from 79% when encoders agree to 27% when two oppose the fused call. Abstaining on disagreement raises balanced accuracy by 7 and 15 points, while retaining 58% and 54% of the cores. Conclusion: ProtoCouncil grounds core-level predictions in exact similarity-based evidence from real, anatomically indexed training cases, which makes its reasoning interpretable. It achieves the highest csPCa AUROC across three cohorts and identifies unreliable predictions through encoder disagreement.
Related papers
- SonoWorld: Towards Interactive Pose-Controllable Ultrasound World Models
- UltraWorld: Learning Interactive Ultrasound World Models from Untracked Clinical Videos with Acoustic Sampling Map
- CoRe-WM: Conditioned Response-State World Modeling for Longitudinal Treatment-Response Prediction in Breast DCE-MRI
- The Predictability Bias of Joint-Embedding Predictive Learning in Medical Imaging
- A small HER2 reader for routine H&E slides: real-world deployment
Publication record
Bibliographic evidence describes where and when a record appeared. It does not validate scientific claims.
First announced by arXiv: 2026-11-09 · Indexed by Pub-AI: 2026-11-09 UTC
ProtoCouncil: Multi-Encoder Anatomical Prototypes for Prostate Cancer Detection in Micro-Ultrasound
First announced by arXiv: 2026-11-09 · Original record ↗
- arxiv
- 2611.09058
Explicit source versions (1)
- v1 · Date not supplied
- Preprint available
Source date: 2026-11-09 · First detected: 2026-11-09 UTC · Historical baseline
Source evidence (1)
- Official evidence ↗ · RSS guid=oai:arXiv.org:2611.09058v1; arxiv:announce_type=new; item pubDate (official metadata)
Abstract · arxiv
Purpose: Micro-ultrasound ($μ$US) is a promising modality for prostate cancer (PCa) detection, but identifying suspicious tissue depends heavily on clinical experience. Models trained on sparse, noisy core-level pathology offer little insight into their decisions and can fail under domain shift. Clinical use needs a traceable decision process and a flag for uncertain predictions. Methods: We introduce ProtoCouncil, in which three foundation encoders (MedSAM, DINOv3 and V-JEPA 2.1) learn case-based prototypes at three anatomical levels: whole gland, needle track and gland sectors. Prototypes are projected onto real training-core regions, and label-free fusion keeps every class logit as a bias plus an exact sum of prototype similarity contributions. Encoder disagreement triggers selective prediction. Results: In five-fold patient-level cross-validation (6,588 cores, 693 patients), ProtoCouncil reaches 82.9% AUROC for clinically significant PCa (csPCa, Grade Group $\geq$3), 4.3 points above the best previously published method. In zero-shot evaluation on two external cohorts, ProtoCouncil achieves higher all-cancer and csPCa AUROC than all previous methods. Nearest same-class exemplars match test-core tumor involvement better than random same-class cases. Externally, csPCa accuracy falls from 79% when encoders agree to 27% when two oppose the fused call. Abstaining on disagreement raises balanced accuracy by 7 and 15 points, while retaining 58% and 54% of the cores. Conclusion: ProtoCouncil grounds core-level predictions in exact similarity-based evidence from real, anatomically indexed training cases, which makes its reasoning interpretable. It achieves the highest csPCa AUROC across three cohorts and identifies unreliable predictions through encoder disagreement.
