ISBI 2026 — Program Notes (post-meeting prep)¶
Status: Scaffold (post-meeting) — meeting concluded April 11, 2026. This page is the candidate shortlist for retrospective bulk extraction into tools/ once IEEE Xplore indexes the 2026 proceedings (expected early May 2026, ~T+3 weeks). Companion to index.md.
Verified meeting metadata¶
- Dates: Wednesday April 8 – Saturday April 11, 2026 (4 days)
- Venue: ExCeL London, London, UK
- Scale: ~1,500 attendees
- Theme: "Translation to Impact!"
- Tagline: "Where machines learn to see the body — and humans learn what comes next"
- Organizers: Joint IEEE Signal Processing Society (SPS) + IEEE Engineering in Medicine and Biology Society (EMBS)
- Proceedings: IEEE Xplore (4-page peer-reviewed conference papers); subset extended to IEEE Transactions on Biomedical Engineering Special Section ("Selected Topics from ISBI 2026") — ~10 invited expanded papers
- Next edition: ISBI 2027 — Lausanne, Switzerland
Why we wait for retrospective extraction (vs live coverage)¶
ISBI 2026 ran April 8–11, overlapping the tail end of AACR 2026 (Apr 17–22 starts ~1 week after ISBI ends, but planning overlap was significant in March–April 2026). Live coverage was deprioritized in favor of AACR + RECOMB + ISMB live builds. ISBI's compact size and post-meeting IEEE Xplore indexing pattern (proceedings drop ~3–4 weeks after the meeting) makes a retrospective extraction tractable and complete — no advantage to a pre-meeting scaffold.
Cancer-relevance filter (the load-bearing axis)¶
Of ISBI 2026's ~400–600 papers, the cancer-relevant subset for this vault is concentrated in four anchor clusters (in descending priority):
1. Pathology / cytology FMs — anchored on the Faisal Mahmood keynote (Harvard Medical School)¶
Mahmood is the lead author / corresponding author on the AACR-corpus pathology FMs CHIEF and UNI. The ISBI 2026 keynote is the likely venue for any methods-side update to that line (UNI2, new tile-encoder pretraining recipe, slide-level aggregation variant). Watch the Mahmood Lab GitHub and HuggingFace through Q2 2026 for the actual artifact drop — the keynote announcement is the signal but the code release is the citable hook.
Crosswalk: any new model goes either (a) as an in-place update to the AACR dossier (incremental work) or (b) as a new ISBI tool page (fundamentally new model). Mirror with the MICCAI 2026 (Sep 27–Oct 1) cousin paper when it lands.
2. RIVA Cervical Cytology Challenge — direct cancer-relevance¶
- Modality: Pap smear cytology (a.k.a. liquid-based cytology slides for cervical cancer precancer detection)
- Dataset: 959 fields of view, 26,158 nucleus-level annotations across Bethesda categories
- Tasks: Nuclei detection + multi-class cell classification (with expert-agreement evaluation as a fairness/inter-rater check)
- Anchor: Direct line to AACR cancer-prevention axis; intersects with CHIEF / UNI as a downstream evaluation use case
- Tool extraction: Top-1 (top-3 if code released) — one page per winning model; the challenge itself is summarized here, not as a separate tool page
3. WBCBench2026 — direct cancer-relevance (hematopathology)¶
- Modality: Peripheral-blood / bone-marrow smear hematopathology
- Tasks: Robust white blood cell classification, with severe class imbalance and fine-grained morphological discrimination — leukemia diagnosis axis
- Evaluation: Patient-level separation enforced to prevent information leakage in test sets (a methodological best-practice flag worth highlighting)
- Anchor: Connects to AACR / ASH hematologic-malignancies dossiers; ASH 2026 (Dec 12–15) clinical-deployment counterpart possible
- Tool extraction: Top-1 winning model + any baseline-beating workshop paper
4. MPI Low Concentration Reconstruction Challenge — direct oncology imaging translation¶
- Modality: Magnetic Particle Imaging (MPI), specifically field-free-line MPI (FFL-MPI) setups
- Tasks: Reconstruction at low tracer concentrations — the bottleneck for clinical translation, with explicit oncology and cardiovascular imaging applications flagged in the challenge description
- Anchor: ISBI is the home venue for MPI methods (MICCAI underweights this modality). MPI is an emerging molecular-imaging modality with potential for tumor-specific tracer visualization
- Tool extraction: Top-1 winning reconstruction algorithm + any methods-paper introducing a new MPI deep-learning prior
5. (Secondary) Foundation models for medical imaging — FAIBI workshop + "Exploring FM in Medical Image Analysis" workshop¶
Both 2026 workshops ran on the closing day (Sat Apr 11). FAIBI 2nd edition co-organized by Jia Wu (MD Anderson Cancer Center) — direct cancer-center authorship signal. Watch for: any oncology-imaging FM update, any pathology FM evaluation paper, any radiology FM with a cancer subgroup analysis.
Keynote roster — full annotation¶
| Speaker | Affiliation | Likely talk axis | Cancer-relevance | Cross-vault link |
|---|---|---|---|---|
| Faisal Mahmood | Harvard Medical School / Brigham & Women's | Pathology FMs; multimodal pathology agents | Direct — CHIEF / UNI line | AACR pathology FMs |
| Polina Golland | MIT CSAIL | MRI reconstruction; image registration; statistical methods | Indirect — methods feed cancer MRI workflows | MICCAI 2026 |
| Bram van Ginneken | Radboud University Medical Center | Diagnostic radiology AI (lung CT, mammography); Grand Challenge platform; clinical translation | Direct — lung cancer screening, breast cancer mammography AI | RSNA 2026 |
| Mauricio Reyes | University of Bern | Trustworthy / explainable medical AI; long-time BraTS organizer | Indirect — brain tumor segmentation legacy | MICCAI 2026 BraTS |
| Greg Slabaugh | Queen Mary University of London | Medical computer vision; local host | Indirect — general medical AI | — |
Challenge slate — full annotation¶
| # | Challenge | Modality | Cancer-direct? | Extract? |
|---|---|---|---|---|
| 1 | Foundation Model Challenge for Ultrasound Image Analysis | Ultrasound (general / fetal-maternal) | No | Only if winner releases a generalist medical-US FM |
| 2 | CXR-LT 2026 | Chest X-ray (long-tail, multi-label) | Indirect (lung cancer in long tail) | If winner addresses oncology-specific subset |
| 3 | Low Concentration Reconstruction in MPI | Magnetic Particle Imaging | Yes — explicit oncology translation | Yes — top-1 + any methods paper |
| 4 | CSV 2026 — Carotid Plaque Segmentation | Carotid ultrasound | No (stroke prevention) | No |
| 5 | Multi-modal Ulcerative Colitis Grading in Endoscopy | Colonoscopy video | Indirect (colon cancer screening axis) | Only if model generalizes to neoplasia detection |
| 6 | FETUS 2026 | Fetal echocardiography | No | No |
| 7 | RIVA Cervical Cytology | Pap smear cytology | Yes — cervical precancer | Yes — top-1 + top-3 if code released |
| 8 | WBCBench2026 | Hematopathology (WBC) | Yes — leukemia | Yes — top-1 + any baseline-beating paper |
Workshop slate — full annotation¶
| Workshop | Day | Cancer-direct? | Extract pages? |
|---|---|---|---|
| Medical Video AI Assessment + UQ | Thu Apr 9 | Indirect (endoscopy generalizes) | Sample only if a named tool released |
| Pediatric Brain Data Analysis | Thu Apr 9 | Indirect (pediatric brain tumors possible) | Sample only |
| Large Models Meet Surgical Data Science | Fri Apr 10 AM | Indirect (surgical-oncology axis) | Sample only |
| Exploring FM in Medical Image Analysis | Fri Apr 10 PM | Yes — explicit FM axis | Yes — extract named-tool papers |
| POCUS-AI | Sat Apr 11 AM | No | No |
| FAIBI 2nd ed. (co-org Jia Wu, MD Anderson) | Sat Apr 11 PM | Yes — cancer-center co-organizer | Yes — extract named-tool papers |
Special-session slate — full annotation¶
| Special Session | Day | Featured | Extract? |
|---|---|---|---|
| Preparing Large-Scale Medical Imaging Data for FM Development | Thu Apr 9 PM | Martin Willemink (Segmed) | If named data-prep tool released |
| Safety and Reliability in Medical Imaging | Thu Apr 9 PM | Gori (Télécom Paris), Zuluaga (EURECOM) | If named uncertainty / safety tool released |
| Digital Twins and Multi-Omics Integration | Fri Apr 10 AM | Valentina Giannini (Turin) | Yes — overlaps AACR multimodal axis |
| Privacy-Aware Federated Learning in Healthcare | Fri Apr 10 AM | Xiaofeng Liu (Yale) | If named federated framework released |
Bulk-extraction plan¶
- Wait for IEEE Xplore indexing (~T+3–4 weeks, early–mid May 2026) — pull the conference paper DOI list. Filter by title/abstract keywords: pathology, cytology, cancer, tumor, oncology, leukemia, lymphoma, carcinoma, MRI reconstruction (selected papers only), MPI, foundation model, multiple instance learning, whole slide, Pap smear, bone marrow, mammography.
- Cross-check against the 4 anchor clusters above. Each cluster yields 2–8 candidate tool pages.
- Verify code/weights availability for each candidate. Skip papers with no released artifact (ISBI 4-page papers without code are weak tool-page candidates).
- Stub tool pages following
tools/index.mdtemplate. Populate the index table. - Cross-link to MICCAI 2026 cousin pages as the September meeting accept-list (already public for early-accept May 7) reveals same-lab follow-ups.
- Track TBME Special Section invitations through Q3 2026; update tool pages with the TBME DOI as the extended version-of-record becomes available.
Estimated yield¶
~10–25 tool pages after the filter. Distribution estimate:
- 3–5 from RIVA + WBCBench2026 + MPI challenges (winners + close runners-up)
- 2–4 from the Mahmood keynote axis (pathology FM updates, if a new artifact drops)
- 3–7 from FAIBI + Exploring-FM-in-Medical-Image-Analysis workshops (named-tool papers)
- 2–5 from main-proceedings oncology-imaging papers (segmentation / MRI / mammography / pathology)
- 0–3 from special-session-announced tools (digital twins, federated learning frameworks)
Cross-vault dependencies¶
- AACR 2026 (dossier index) — Mahmood keynote work crosswalks here first. Update CHIEF / UNI / GigaPath dossiers in-place for incremental work.
- MICCAI 2026 (index) — same author pool largely; ISBI April → MICCAI September cousin pairing is the rule. Wait for the MICCAI early-accept list (already posted May 7, 2026) and the final-accept list (Jun 12) to identify cousin papers.
- RSNA 2026 (index) — Bram van Ginneken keynote work crosswalks to RSNA's clinical-deployment side. Lung CT and mammography AI specifically.
- ASH 2026 (Dec 12–15) — WBCBench2026 leukemia-classification winners may resurface as ASH clinical-deployment readouts.
- Conferences timeline (timeline.md) — ISBI sits in early April, ~1 week before AACR.