2026 ASH®
Annual Meeting & Exposition December 12-15, 2026 ▪ New Orleans, Louisiana

Paper Details

Abstracts, authors, affiliations, and disclosures.

Monday, December 14, 2026
06:00 p.m. - 08:00 p.m.
MCCNO - Hall F
Technology and Procedures, Assays, Imaging, Molecular Testing, Newborn Screening, Omics Technologies, Pathology, Serologic Tests, Bioinformatics, Computational Biology, Emerging Technologies, Profiling, Treatment Considerations, Combination Therapy, Therapy Sequence, Biological Therapies, Antibody Therapy, Transplantation, Bispecific Antibody Therapy, (CAR)-T Cell Therapies, Monoclonal Antibody Therapy, Non-Biological Therapies, Chemotherapy, Surgical, Adverse Events, Artificial Intelligence (AI), Computer Vision, Deep Learning, Generative AI, Machine Learning, Natural Language Processing, Biological Processes, Pathogenesis, CHIP, Genomics, Hematopoiesis, Immune Mechanism, Metabolism, Molecular Biology, Multi-Systemic Interactions, Clinical Practice (Health Services and Quality), Clinical Guidelines, Diseases, Hemoglobinopathies, Sickle Cell Disease, Immunology, GVHD, Lymphoid Malignancies, Lymphoid Leukemias, Lymphomas, Plasma Cell Disorders, Myeloid Malignancies, Acute Myeloid Malignancies, Chronic Myeloid Malignancies, Measurable Residual Disease, ALL, CLL, Non-Hodgkin Lymphoma, B Cell Lymphoma, Indolent Lymphoma, Aggressive Lymphoma, Research, Fundamental Science, AML, APL, Translational Research, MDS, MPN, CMML, Clinical Research, Clinical Trials, Health Outcomes Research, Patient-Reported Outcomes, Real-World Evidence, Registries, Survivorship, Study Population, Human, Adult, Elderly, Pediatric
PAPER

6099 Lymphovision morphology prediction on diagnostic biopsies identifies a subset of follicular lymphoma at high risk of histologic transformation

Jansen N Seheult, MBBCh, BAO, MSc, MS, MD, Wenchao Han, PhD2, Reyhan Keser, PhD3, Ayo Samuel Falade, MD, MBA4, Steven Huang5, Yurim Lee, PhD6, Raphael Mwangi, MS7, Janek S. Walker, PhD8, Joseph Peter Novak, MS9, Thomas Tavolara, PhD10, Aishwarya Ravindran, MBBS11, Thomas Witzig, MD12, Jose C Villasboas, MD13, Cheryl Willman, MD14, Philippe Armand, MD15, Stephen Maxted Ansell, MD, PhD16, Anne J. Novak, PhD17, James R. Cerhan, MD, PhD18, Patrizia Mondello, MD, PhD19, Rebecca L King, MD20, Matthew J. Maurer, Sc.D., M.S.21, Eric Hsi, M.D.22
1 Mayo Clinic, Rochester, Minnesota, United States
2 Mayo Clinic, Rochester, Minnesota, United States
3 Mayo Clinic, Rochester, Minnesota, United States
4 Mayo Clinic, Rochester, Minnesota, United States
5 Mayo Clinic, Rochester, Minnesota, United States
6 Mayo Clinic, Rochester, Minnesota, United States
7 Mayo Clinic, Mayo Clinic, Minnesota, United States
8 Mayo Clinic, Rochester, Minnesota, United States
9 Mayo Clinic, Mayo Clinic, Minnesota, United States
10 Mayo Clinic, Rochester, Minnesota, United States
11 Mayo Clinic, Rochester, Minnesota, United States
12 Mayo Clinic, Rochester, Minnesota, United States
13 Mayo Clinic, Rochester, Minnesota, United States
14 Mayo Clinic, Rochester, Minnesota, United States
15 Mayo Clinic, Rochester, Minnesota, United States, Mayo Clinic, Rochester, Minnesota, United States
16 Mayo Clinic, Rochester, Minnesota, United States, Mayo Clinic, Rochester, Minnesota, United States
17 Mayo Clinic, Mayo Clinic, Minnesota, United States
18 Mayo Clinic, Mayo Clinic, Minnesota, United States, Mayo Clinic, Rochester, Minnesota, United States
19 Mayo Clinic, Rochester, Minnesota, United States
20 Mayo Clinic, Rochester, Minnesota, United States
21 Mayo Clinic, Mayo Clinic, Minnesota, United States
22 Mayo Clinic, Rochester, Minnesota, United States

Background: Histologic transformation (HT) is a major driver of excess mortality in follicular lymphoma (FL), yet no validated morphologic predictor of HT exists at diagnosis. FLIPI and histologic grade discriminate poorly. LymphoVision (LV), a lymphoma-specialized pathology foundation model (Seheult et al, Blood 2025;146 Suppl 1:117), may assign a non-FL diagnosis to cases diagnosed as FL in routine practice. We examined whether such discordant calls are associated with HT risk.

Methods: We identified diagnostic specimens from 262 patients with newly diagnosed grade 1-3A FL who were prospectively enrolled in the Molecular Epidemiology Resource (MER) between 2002 and 2024. All diagnoses were reviewed by an expert hematopathologist and managed clinically as FL; eligibility was based on this diagnosis of record, with follow-up per MER protocol.

We applied LV, a DINOv2 ViT-G/14 image foundation model pretrained on 37,406 H&E slides from 12,919 lymphoproliferative specimens. A frozen multiple-instance-learning head generated a lymphoma classification call [diffuse large B-cell (DLBCL; training class included high-grade B-cell lymphomas), FL, mantle cell (MCL), marginal zone, classical Hodgkin, benign/reactive]. The MCL training corpus was enriched for aggressive disease (~50% high Ki-67, ~30% blastoid or pleomorphic morphology). No retraining or calibration was performed; the present cohort had no overlap with model training or validation data. Non-FL calls (LV-nonFL) triggered hematopathologist re-review, unblinded to model prediction. Time to HT was analyzed with death without HT as a competing risk (cumulative incidence, Gray's test, Fine-Gray regression); event-free survival at 24 months (EFS24) was a secondary endpoint.

Results: LV assigned 242 cases (92.4%) to FL and 20 (7.6%) to LV-nonFL (16 DLBCL, 4 MCL). At baseline, LV-nonFL cases more often had elevated LDH (45% vs 16%, p=0.004) but did not differ in high-risk FLIPI24 (42% vs 28%, p=0.29) or grade 3A histology (15% vs 5%, p=0.10); they were more often treated with first-line immunochemotherapy (IC, 60% vs 31%, p=0.013) and less often observed (20% vs 45%, p=0.036).

At a median follow-up of 99.5 months (IQR 60.0-156.8), 39 HTs occurred (14.9%). Five-year cumulative incidence was 35% (95% CI 19-65) for LV-nonFL vs 9% (6-13) for FL-predicted cases (subdistribution HR [sHR] 2.95, 1.25-7.01, p=0.014); 2-year cumulative HT incidence was 20% vs 5%. HT occurred in 5 of 16 (31%) DLBCL and 2 of 4 (50%) LV-MCL calls. The association was consistent in patients not receiving R-CHOP (5-year 38% vs 8%, p=0.022) and directionally consistent in those who did (29% vs 11%, p=0.25), was consistent in grade 1-2 FL (35% vs 8%, p=0.020), and independent of FLIPI24 in multivariable Fine-Gray regression (sHR 2.41, 1.02-5.71, p=0.045).

Among patients initially observed, 2 of 4 (50%) LV-nonFL transformed vs 16 of 108 (14.8%) FL-predicted. EFS24 failure was also more frequent in LV-nonFL cases (45% vs 22%, p=0.026), though not significantly different within first-line IC alone (42% vs 21%, p=0.15). On re-review, 3 LV-nonFL cases were grade 3A FL with large-cell or blastoid cytology, and 3 of 4 MCL calls were grade 1-2 FL with high Ki-67. In 2 cases, the original report favored FL without stating it definitively; findings were unchanged when these 2 were excluded (6 of 18 transformed). Six of 20 cases showed crush artifact and/or sclerosis. Among the 14 cases without high-grade, blastoid and/or high-proliferation (HG/HP) morphology on re-review, 4 (29%) transformed and 8 (57%) failed EFS24, versus 3 (50%) and 1 (17%) of the 6 with these features.

Conclusions: In grade 1-3A FL, an LV-nonFL prediction identified a subgroup with an approximately 4-fold higher cumulative incidence of HT, independent of FLIPI24. LV-nonFL patients received more intensive therapy yet transformed more often, arguing against undertreatment. Pathologist re-review indicated a heterogeneous group: HG/HP morphology, calls likely driven by diffuse architecture, fibrosis, or crush artifact, and low-grade cytology; HT occurred in all three, and EFS24 failure was concentrated in cases without HG/HP morphology. Because the MCL training corpus was enriched for HG/HP morphology, an MCL call on grade 1-2 FL may encode high proliferative index rather than mantle-cell lineage. Limitations include single-institution design, 20 discordant cases with 7 events and unblinded re-review. External validation is in progress.

Jansen N Seheult, MBBCh, BAO, MSc, MS, MD: No relevant conflicts of interest to declare.