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

Paper Details

Abstracts, authors, affiliations, and disclosures.

Saturday, December 12, 2026
05:30 p.m. - 07:30 p.m.
MCCNO - Hall F
Access, Representation, Population Health, Technology and Procedures, Imaging, Molecular Testing, Omics Technologies, Pathology, Bioinformatics, Blood Banking, Computational Biology, Emerging Technologies, Profiling, Treatment Considerations, Drug Development, Supportive Care, Therapy Sequence, Biological Therapies, Chimeric Antigen Receptor, (CAR)-T Cell Therapies, Women’s Health, Workforce, Adverse Events, Artificial Intelligence (AI), Computer Vision, Deep Learning, Generative AI, Machine Learning, Natural Language Processing, Biological Processes, Apoptosis, Genomics, Hematopoiesis, Molecular Biology, Clinical Practice (Health Services and Quality), Clinical Guidelines, Quality Improvement, Quality of Care, Systems-Based Hematology, Diseases, Bleeding and Clotting, Thromboembolism, Bone Marrow Failure Syndromes, Inherited Marrow Failure Syndromes, Paroxysmal Nocturnal Hemoglobinuria, Lymphoid Malignancies, Lymphoid Leukemias, Lymphomas, Plasma Cell Disorders, Myeloid Malignancies, Acute Myeloid Malignancies, Chronic Myeloid Malignancies, Metabolic Disorders, Iron Deficiency, Measurable Residual Disease, Non-Hodgkin Lymphoma, B Cell Lymphoma, Aggressive Lymphoma, Research, Fundamental Science, AML, APL, Translational Research, MDS, MPN, Clinical Research, Clinical Trials, Real-World Evidence, Study Population, Human, Adult, Pediatric
PAPER

2588 FlowER: Computational classification of phenotypically aberrant cells in acute myeloid leukemia measurable residual disease by flow cytometry

Ansh Kumar, Jake Silberg2, Meindert Niemeijer, PhD3, Bradley Patay, MD4, Robert Andrews5, Jeffrey Miller, PhD6
1 Invivoscribe, Inc., San Diego, California, United States
2 Invivoscribe, Inc., San Diego, California, United States
3 Invivoscribe, Inc., San Diego, California, United States
4 Invivoscribe, Inc., San Diego, California, United States
5 Invivoscribe, Inc., San Diego, California, United States
6 Invivoscribe, Inc., San Diego, California, United States

Background: Multiparameter flow cytometry (MFC) is widely used to assess measurable residual disease (MRD) in acute myeloid leukemia (AML), but interpretation often depends on manual gating and expert recognition of abnormal immunophenotypes. Interpretation can vary across operators and gating strategies, especially when leukemic populations are heterogeneous or overlap with normal maturation. An objective, reference-based method for identifying phenotypically aberrant cells could improve reproducibility while preserving the resolution of MFC. We developed FlowER, a reference-based framework that classifies individual cells according to their deviation from normal reference populations without per-cell supervision.

Methods: FlowER compares each query cell with a curated library of normal bone marrow specimens. For each cell, it assesses how consistently its phenotype differs from those observed across the reference library, whether that phenotype is enriched in the query specimen relative to the references, and whether the deviation is supported by the cell’s local phenotypic neighborhood. These measures are combined to assign each cell a binary abnormality flag, which preserves the identity of each cell and identifies populations that depart from normal hematopoietic organization. The method does not require predefined leukemia-associated immunophenotypes or supervised training on positive-cell labels.

FlowER was evaluated using viable CD45-dim cells from one tube of a 12-color AML MFC panel across 257 bone marrow specimens, including 163 with hematopathologist-gated aberrant populations and 94 MRD-negative patient specimens. FlowER’s classifications were compared directly with clinical cell-level gate membership, and, separately, predicted abnormal-cell burden was calculated as the number of CD45-dim cells classified as aberrant divided by the total viable-cell count and compared with clinically gated burden. The method’s performance was assessed using precision, sensitivity, specificity, accuracy, F1 score, and a Spearman rank correlation between predicted and clinically gated burden.

Results: Across 21,767,001 cells from 257 specimens, FlowER achieved a cell-pooled precision of 0.881, sensitivity of 0.873, accuracy of 0.844, and F1 score of 0.877 for identifying hematopathologist-gated aberrant cells. Among the 163 specimens containing clinically gated abnormal populations, pooled precision was 0.888, sensitivity was 0.873, and F1 score was 0.881; median per-specimen precision, sensitivity, and F1 score were 0.704, 0.832, and 0.620, respectively.

FlowER-derived abnormal-cell burden was strongly correlated with clinically gated burden across the entire cohort (Spearman ρ=0.823, p=1.11×10-64), demonstrating agreement in the relative extent of abnormal populations in addition to cell-level performance. Among the 94 negative patient specimens, pooled specificity was 0.934, corresponding to a pooled false-positive fraction of 0.066. The median specimen-level false-positive fraction was 0.030.

Conclusions: FlowER provides objective, reference-based classification of phenotypically aberrant cells in AML MFC data without disease-specific manual gates or any prior specification of disease-associated phenotypes. The framework, across 257 specimens and over 21 million cells, achieved a high sensitivity and F1 score, strong concordance between predicted and clinically gated abnormal-cell burden, and a high specificity in negative specimens. By converting multiple independent measures of deviation from normal into a cell-level decision, FlowER offers a reproducible and scalable approach for reducing subjectivity and inter-operator variability in MFC interpretation while remaining compatible with established hematopathology workflows. 

Ansh Kumar: Invivoscribe, Inc. - Patents & Royalties; Invivoscribe, Inc. - Consultancy (Includes expert testimony)