Dalland, Joanna

Abstract Background: Recent advances in computer vision have given rise to image foundation models (FMs), large-scale neural networks pre-trained on millions of unlabeled histopathology patches, which can learn generalizable representations of tissue morphology. In many cancers, these FMs have predicted genetic mutations and prognoses directly from histology, suggesting that morphology encodes “hidden” molecular information. In lymphoma, this field is nascent but growing.General-purpose histopathology FMs have limited representation of lymphoid morphology. We developed and evaluated LymphoVision, a lymphoma-specialized image foundation model pretrained with self-supervised learning, to extract clinically meaningful features from routine hematoxylin and eosin (H&E) histology. Methods: LymphoVision was pre-trained using knowledge distillation and consistency (DINOv2) on over 37 million multi-resolution patches (5x, 10x, 20x, 40x) sampled using a density-aware strategy from 31,211 H&E whole slide images (WSIs) from 9,155 archival lymphoma and reactive lymphoproliferative cases. To ensure diverse, representative samples, our patch sampling method uses an efficient graph-based clustering algorithm and prioritizes morphologically informative regions by clustering patch embeddings and performing density sampling within each cluster. The pretrained encoder was evaluated using a clustering-constrained attention multiple instance learning (MIL) framework on three diagnostic tasks: (1) cell-of-origin (COO) classification in diffuse large B-cell lymphoma (DLBCL), distinguishing germinal center B-cell (GCB) from non-GCB subtypes based on the Hans immunohistochemistry algorithm; (2) follicular lymphoma (FL) grading, differentiating low-grade (grades 1–2) from high-grade (grade 3); and (3) multiclass lymphoma subtype classification across six categories: DLBCL, FL, marginal zone lymphoma (MZL), mantle cell lymphoma (MCL), classical Hodgkin lymphoma (CHL), and benign/reactive lymphoid conditions. WSI-level labels were derived from pathology reports, and none of the evaluation cases were included in pretraining. Models used a ViT-Giant LymphoVision backbone to generate 20x patch embeddings, followed by MIL using 50% training data and 2-fold cross-validation (25% validation, 25% test). Performance was assessed using area under the receiver operating characteristic curve (AUC), accuracy, recall (sensitivity) and F1 score. Results: For DLBCL COO classification (total n=358 cases), LymphoVision achieved a mean test AUC of 0.93 (range 0.92–0.94), accuracy of 84.5% and F1 of 0.85 across two independent test folds (n=91 per fold, 50% GCB). These results outperformed general-purpose FMs on the same task and cases, including Virchow-v2 (AUC 0.85) and UNI-v2 (AUC 0.88), despite those models being pretrained on 10–100 times more images. For FL grading (total n=660 cases), the test set in each fold included 166 cases (18% grade 3), yielding a mean test AUC of 0.88 (range 0.87–0.90), accuracy of 89.2% and F1 of 0.80. In multiclass lymphoma subtype classification (total n=1,176 cases), the test set included 296 cases per fold. LymphoVision achieved a mean test AUC of 0.98 (range 0.97-0.99), mean test accuracy of 87.1% and mean F1 of 0.84. Class-wise recall was 92.0% for DLBCL (n=112), 88.2% for FL (n=72), 77.2% for MZL (n=29), 70.8% for MCL (n=35), 87.5% for CHL (n=24), and 95.8% for benign/reactive lymphoid conditions (n=24). Conclusion: LymphoVision is a lymphoma-specialized FM trained using self-supervision that achieves high diagnostic accuracy across multiple clinically meaningful tasks using weak supervision and standard H&E images, suggesting the encoder has learned fine-grained lymphoma morphology. Based on historical benchmarks from the 1997 International Lymphoma Study Group classification project, expert hematopathologists achieved diagnostic agreement rates of 55% to 84% using H&E morphology alone for the non-Hodgkin lymphoma (NHL) entities included in our multiclass task. LymphoVision approached or exceeded these benchmarks. These findings underscore the value of our approach using disease-specialized pretraining and intelligent patch sampling for data-efficient histopathology learning. LymphoVision has the potential to augment diagnostics, enable scalable AI-powered classification in real-world workflows, and serve as a foundation for prognostic and theragnostic applications. External validation and multimodal integration are ongoing.

Abstract Objective The aim of this study was to evaluate a large cohort of gallbladder amyloid cases to determine clinical and morphologic features. Methods Cholecystectomy specimens (N = 118) typed using proteomics-based techniques between 2008 and 2023 were identified. Clinical and morphologic features were reviewed. Results Six amyloid types were identified: ATTR (n = 63, 53.4%), AL (n = 46, 39.0%), AA (n = 4, 3.4%), AApoA1 (n = 2, 1.7%), ALECT2 (n = 2, 1.7%), and AEFEMP1 (n = 1, 0.8%). Amyloidogenic mutations were detected in 3 ATTR cases and 2 AApoA1 cases. Morphologic review (n = 26) revealed perimuscular vessel involvement in all cases. Amyloidosis was an unexpected diagnosis first made on the cholecystectomy specimen in half of the patients with clinical information (n = 10). All 9 patients with follow-up had evidence of systemic disease. In 2 patients, cholecystic involvement was initially missed and only retrospectively identified after the diagnosis of cardiac amyloidosis. Conclusions In patients with clinical data, amyloidosis was often unexpected, the gallbladder was commonly the first tissue sampled with amyloidosis, and all patients had systemic disease. Thorough review of cholecystectomy specimens with careful inspection of perimuscular vessels, coupled with a low threshold for ordering Congo red stain in elderly individuals and amyloid typing using a robust method such as proteomics, can prevent a delay in amyloid diagnosis and management.

Abstract Introduction/Objective Distinguishing follicular lymphoma (FL) from marginal zone lymphoma (MZL) can be challenging, as the morphology and immunophenotype can overlap particularly in CD10-negative cases or those with marginal zone differentiation (MZD). MEF2B (Myocyte enhancer binding factor B) is a transcription factor that controls BCL6 expression in germinal center (GC) B-cells. Previous studies report MEF2B to have a high sensitivity and specificity for FL compared to other GC markers. We investigated the expression of MEF2B in low grade B cell lymphomas (LGBCL) to evaluate its utility in classifying diagnostically challenging cases and FL with MZD. Methods/Case Report Our archives were searched from 2005-2023 to identify excisional biopsies of LGBCL with MZD, including (FL), splenic MZL (SMZL), MZL of mucosa-associated lymphoid tissue (MALT), nodal MZL (NMZL) and difficult-to-classify (DTC) cases with a differential diagnosis of MZL and FL. MEF2B immunohistochemistry was performed retrospectively (Atlas Antibodies). Two hematopathologists assigned a pre-and post-MEF2B diagnosis to DTC cases. Results (if a Case Study enter NA) 57 cases were included: 21 FL (12 with MZD), 11 SMZL, 13 MALT, 4 NMZL, and 8 DTC cases. 19/21 FL cases (90%) were positive for MEF2B, including 11/12 (92%) cases with MZD. However, MEF2B expression was restricted to the follicles and negative in MZD areas in 7/11 cases (64%), while 4 cases showed partial weak expression in the areas of MZD. 0/11 SMZL (0%), 0/13 MALT (0%), and 0/4 (0%) NMZL were positive for MEF2B. 6/8 DTC cases (75%) were positive for MEF2B. Within DTC cases, retrospective inclusion of MEF2B results helped to render a more definitive diagnosis in 5 cases (2 negative/3 positive). Conclusion MEF2B immunohistochemistry can aid in distinguishing FL from MZL in challenging cases. However, it is often negative in areas of FL with MZD so care must be taken on interpretation in small biopsies.

The spleen is a commonly encountered specimen in surgical pathology. However, little is known about the incidence, morphologic pattern, and clinical features of spleens involved by amyloidosis. We retrospectively identified 69 spleen amyloid cases typed using a proteomics-based method between 2008 and 2020. The frequency of amyloid types, clinicopathologic features, and distribution of amyloid deposits were assessed. Four amyloid types were detected: immunoglobulin light chain (AL) (N=30; 43.5%); leukocyte chemotactic factor 2 amyloidosis (ALECT2) (N=30; 43.5%); amyloid A (AA) (N=8; 11.6%); and fibrinogen alpha (AFib) (N=1; 1.4%). The splenic amyloid showed 5 distinct distribution patterns: (1) diffuse pattern, exhibited by most AL cases; (2) red pulp pattern, exhibited by most ALECT2 cases; (3) multinodular pattern, seen in subsets of AA and AL-kappa cases; (4) mass-forming pattern, seen in the AFib case; and (5) vascular only, seen in a subset of AA cases. Atraumatic splenic rupture was the most common reason for splenectomy in AL cases, while most ALECT2 spleens were removed incidentally during an unrelated abdominal surgery. Splenomegaly was significantly more common in AA spleens than in AL or ALECT2 spleens and was often the reason for splenectomy in this group. In conclusion, splenic amyloid may be underrecognized as it is often an incidental finding. Although, as expected, many of the spleens were involved by AL amyloidosis, ALECT2 emerged as another common spleen amyloid type. Although the spleen amyloid types exhibited characteristic distribution patterns, proteomics-based typing is warranted as some morphologic overlap still exists. Awareness of ALECT2 as a major spleen amyloid type is important for appropriate diagnostic workup and patient management.

Abstract The World Health Organization category of myeloid/lymphoid neoplasms with eosinophilia and PDGFRA rearrangements is composed of a heterogeneous group of neoplasms that can present as a myeloproliferative neoplasm, acute myeloid leukemia, myeloid sarcoma, or lymphoblastic leukemia/lymphoma. The overall outcome of these neoplasms is favorable with imatinib therapy. Herein, we describe an adult female patient with a myeloid neoplasm accompanied by eosinophilia and a novel USP25::PDGFRA gene fusion.

AbstractPlasma cell neoplasms (PCN) and mantle cell lymphoma (MCL) can both harbor t(11;14)(q13;q32) (CCND1/IGH), usually resulting in cyclin D1 overexpression. In some cases, particularly at low levels of disease, it can be morphologically challenging to distinguish between these entities in the bone marrow (BM) since PCN with t(11;14) are often CD20‐positive with lymphoplasmacytic cytology, while MCL can rarely have plasmacytic differentiation. We compared the difference in CCND1/IGH by fluorescence in situ hybridization (FISH) in PCN and MCL to evaluate for possible differentiating characteristics. We identified 326 cases of MCL with t(11;14) and 279 cases of PCN with t(11;14) from either formalin‐fixed, paraffin‐embedded tissue or fresh BM specimens. The “typical,” balanced CCND1/IGH FISH signal pattern was defined as three total CCND1 signals, three total IGH signals, and two total fusion signals. Any deviation from the “typical” pattern was defined as an “atypical” pattern, which was further stratified into “gain of fusion” vs “complex” patterns. There was a significantly higher proportion of cases that showed an atypical FISH pattern in PCN compared with MCL (53% vs 27%, P < .0001). There was also a significantly higher proportion of cases that showed a complex FISH pattern in PCN compared with MCL (47% vs 17%, P < .0001). We confirmed these findings using mate‐pair sequencing of 25 PCN and MCL samples. PCN more often have a complex CCND1/IGH FISH pattern compared with MCL, suggesting possible differences in the genomic mechanisms underlying these rearrangements in plasma cells compared with B cells.

AbstractObjectivesMyeloid sarcoma with CBFB-MYH11 fusion may be enriched in abdominal sites. The clinicopathologic features of 11 cases are reported.MethodsWe collected clinical features, morphology, immunophenotype, and bone marrow (BM) involvement of myeloid sarcoma cases with CBFB-MYH11 fusion.ResultsEleven of 29 total myeloid sarcoma cases were CBFB-MYH11 positive and all 11 involved abdominal sites. The blastic infiltrate was associated with eosinophils in four of 11 cases and plasmacytoid dendritic cell (pDC) nodules in four of six cases. CD34, CD117, and myeloperoxidase were expressed in eight of nine, 10 of 10, and 10 of 10 cases, respectively. Ten of 10 cases showed no BM involvement.ConclusionsOur current series, combined with a literature review, identifies a compelling series of 31 (94%) of 33 cases of myeloid sarcoma with CBFB-MYH11 fusion showing a marked predilection for abdominal sites. In addition, the lack of obvious associated eosinophils, presence of pDC nodules, and lack of concurrent BM involvement suggest that “myeloid sarcoma with CBFB-MYH11 fusion” may represent a unique phenomenon.