Major WHO-HAEM5/ICC categories of myeloid neoplasms | Immunophenotypic characterization/alteration | Significance | Citation no. | Literature review |
---|---|---|---|---|
reference by author/year/journal | ||||
MDS | High concordance of aberrant findings with cytomorphology; >3% CD34+ myeloid progenitors highly associated with MDS | Shameli et al. (2021), Cytometry B Clin Cytom 100 | ||
Chan et al. (2023), Cytometry B Clin Cytom 104 | ||||
Subira et al. (2021), Ann Hematol 100 | ||||
Refer to Table 2 for additional references | ||||
MPN | Myeloid progenitors (quantification; aberrant immunophenotype) Alterations in maturing myelomonocytic cells; use of “monocyte assay” | Distinction between MPN with monocytosis & CMML; quantification & characterization of blasts Detection of evolving lymphoblastic crisis in CML; distinction between MPN-AP & de novo acute leukemia | Ouang et al. (2015), Cytometry B Clin Cytom 88 | |
Bassan et al. (2022), Med Oncol 39 | ||||
Kern et al. (2013), Cytometry B Clin Cytom 84 | ||||
Mannelli et al. (2022), Am J Hematol 97 | ||||
Guglielmelli et al. (2017), Blood 129 | ||||
Jeryczynski et al. (2017), Am J Hematol 92 | ||||
Bardet et al. (2015), Haematologica 100 | ||||
El Rassi et al. (2015), Cancer 121 | ||||
Chan et al. (2023), Cytometry B Clin Cytom 104 | ||||
MDS/MPN | MDS-type immunophenotypic alterations; quantification & characterization of myeloid progenitors; immunophenotypic characterization of monocytic subsets | Abnormal partitioning of monocytic subsets as diagnostic criterion in CMML (versus other myeloid neoplasms with monocytosis versus reactive monocytosis) | [26] | Patnaik et al. (2017), Blood Cancer J 7 |
Wagner-Ballon et al. (2023), Cytometry B Clin Cytom 104 | ||||
Solary et al. (2020), Best Pract Res Clin Haematol 33 | ||||
Hudson et al. (2018), Leuk Res 73 | ||||
Kern et al. (2013), Cytometry B Clin Cytom 84 | ||||
Shameli et al. (2021), Cytometry B Clin Cytom 100 | ||||
Li et al. (2021), Am J Clin Pathol 156 | ||||
Huang et al. (2016), J Clin Pathol 69 | ||||
Maioli et al. (2016), Leuk Lymphoma 57 | ||||
Cargo et al. (2019), Blood 133 | ||||
M/LN-eo-TK | Broad myeloid/lymphoid FCM panels | Identification of small subpopulations, not apparent by morphology (immunohistochemistry); lineage infidelity; aberrant mast cells | A large number of single case reports & literature reviews on specific subcategories of M/LN-eo-TK (recently reviewed in [24]) illustrate how FCM can assist in establishing the diagnosis | |
Mastocytosis | Aberrant immunophenotype, including expression of CD2, CD25, and/or CD30 | Included as diagnostic criterion; detection of associated myeloid neoplasm or concurrent lymphoid or plasma cell neoplasm | Pardanani et al. (2009), Blood 114 | |
Pardanani et al. (2016), Leukemia 30 | ||||
Morgado et al. (2013), Histopathology 23 | ||||
Russano de Paiva et al. (2018), Medicine (Baltimore) 97 | ||||
AML defining genetic abnormalities; AML defined by differentiation (WHO-HAEM5)/AML, NOS (ICC) | Phenotype–genotype correlation (recently reviewed in [17]); monocytic/megakaryocytic differentiation; minimal differentiation | Rapid diagnosis in APL; guidance for additional testing; MRD monitoring; FCIP required for AML lacking defining genetic abnormalities (update on differentiation markers in WHO-HAEM5/ICC) | Merati et al. (2021), Front Oncol 11 | |
Xiao et al. (2021), Blood 137 | ||||
Wang et al. (2022), Cancers 14 | ||||
BPDCN | Expression of CD123 and one other pDC marker in addition to CD4 and/or CD56; or expression of any three pDC markers & absence of all expected negative markers | Immunophenotypic diagnostic criteria | Wang et al. (2022), Cancers 14 | |
Khoury et al. (2018), Curr Hematol Malignancy Reports 13 | ||||
Deotare et al. (2016), Am J Hematol 91 | ||||
Wang et al. (2021), Haematologica 106 |
Reference | Type of disease studied/task | Type of AI tool/application | Significance |
---|---|---|---|
Author/year/journal | |||
General review | |||
Duetz et al. (2020), Curr Opin Oncol 32 | Computational analysis of FCM data in hematological malignancies; literature review | Two main types of computational methods; dimensionality reduction & clustering | Increase ease of use, objectivity & accuracy of FCM data analysis; integration with digital pathology approaches |
Bene et al. (2021); [2] | AI applications with focus on hematological neoplasms, including clinical examples | Various AI tools for automated, unsupervised FCM analysis; FlowSOM | Integration of FlowSOM & existing FCM software programs (e.g., Kaluza, Beckman Coulter) links unsupervised & supervised analysis |
Shouval et al. (2021), Br J Haematology 192 | AI applications in clinical hematology including literature review | Various references on supervised ML studies in hematology | Provides tools & guidance for understanding ML and its applications |
Bone marrow and peripheral blood | |||
Lacombe et al. (2019), HemaSphere 3 | Normal or diseased BM subsets | FlowSOM; Kaluza software program | Objective delineation of BM differentiation pathways |
Zhang et al. (2020), Am J Clin Pathol 153 | FCM PB screening for hematologic malignancy | ML tool with clinical information & laboratory values as input data | Decision tree model for triaging PB FCM specimen; not considered appropriate for screening of a general population |
Flores-Montero et al. (2019), J Immunol Methods | Automated identification of PB lymphocyte subsets for chronic lymphoproliferative disorders | EuroFlow Lymphoid Screening Tube (LST) data base | Reliable & reproducible tool for fast identification of normal vs pathological B and T/NK lymphocytes |
Lymphoid neoplasms | |||
Scheuermann et al. (2017), Clin Lab Med 37 | FCM based identification of diagnostic cell populations in CLL patient samples | FLOCK-(Flow clustering without K) based computational pipeline (publically available for open use, http://www.immport.org) | Clinical validation of computational approaches for use in the clinical laboratory |
Moraes et al. (2019), Comput Methods Programs Biomed 178 | FCM-based, automated classification of mature lymphoid neoplasms | Decision-tree approach for the differential diagnosis using logistic function nodes | Validated scheme in diagnostic samples |
Gaidano et al. (2020), Cancers 12 | FCM-based classification of mature lymphoid neoplasms | ML based on manual FCM analysis of clinical cases from a database | High accuracy for common clinicopathological entities |
Salama et al. (2022), Cancers 14 | FCM MRD in CLL | Deep neuronal network for MRD detection; “human-in-the-loop” AI approach | High accuracy in CLL MRD detection; provides framework for testing in other hematologic disorders |
Simonson et al. (2021), Am J Clin Pathol 156 | FCM ML in classic Hodgkin Lymphoma | CNN for detecting cHL using FCM data (two-dimensional histograms) | New ML algorithm with focus on explainability & visualization (Shapley additive explanation value) |
Nanaa et al. (2021), Pathology 53 | Literature review; AI application in the diagnostics of leukemia & lymphoma | AI algorithms applied to digital morphology and FCM | High accuracy of AI tools in diagnostic hematopathology |
Zhao et al. (2020), Cytometry Part A, 97A | FCM classification of mature B cell neoplasms | Transformation of FCM raw data into a single image file (SOM), further analyzed by CNN for pattern recognition | SOM-CNN-based classification method able to differentiate eight B‑NHL subtypes & normal controls with high accuracy |
Nguyen et al. (2023), Br J Hematol 00 | FCM CLL MRD | Flow SOM | Feasibility & value of automated FCM analysis in the clinical laboratory |
Plasma cell disorders | |||
Sanoja-Flores et al. (2018), Blood Cancer J 8 | Characterization of MGUS & PCM | Next-generation FCM approach on circulating plasma cells | Correlation with diagnostic and prognostic disease categories |
Clichet et al. (2022), Br J Hematol 196 | FCM classification of plasma cell dyscrasias (MGUS, SPCM, PCM) | Immunophenotypic profile analysis (FCM) based on a gradient boosting machine (GBM) algorithm using seven FCM parameters | Expression of CD27 & CD38 was found crucial to discriminate MGUS from MM |
Flores-Montero et al. (2017), Leukemia 31 | MRD plasma cell myeloma | EuroFlow-based NGS FCM; standardized approach for MRD detection | Improved sensitivity for MRD detection; prognostic value; ready for implementation in routine diagnostics |
Myelodysplastic syndrome | |||
Barreau et al. (2019), Cytom B Clin Cytom 98 | Evaluation of maturation of granulocytes & monocytes | Manual expert analyzed FCM score | Improvement of accuracy of FCM diagnosis |
Duetz et al. (2021), Cytometry 99 | Computational workflow for MDS diagnosis; distinction between MDS and non-neoplastic cytopenias | FlowSOM, random forest (ML qualifier) | Workflow outperformed the conventional, expert analyzed FCM scores with respect to accuracy, objectivity, and turn-around time |
Clichet et al. (2023), Haematologica, online, ahead of print | FCM based model to predict MDS | ML model based on FCM parameters selected by Boruta algorithm | Improved the sensitivity of the Ogata score; used both in low & high risk MDS |
Porwit et al. (2022), Cytometry B Clin Cytom 102 | FCM analysis of normal BM & BM from MDS patients targeting erythropoiesis | FlowSOM Identification of 6 subpopulations of erythropoietic precursors in normal BM & additional 18 subsets in MDS | Unsupervised clustering analysis of FCM data disclosed subtle alterations not detectable by FCM supervised analysis |
Acute leukemia and MRD | |||
Monaghan et al. (2022), Am J Clin Pathol 157 | Assessment of BM in unclear cytopenia and/or AL | ML model based on a 37-parameter FCM panel for AL diagnosis & classification | Use of three parameters including light scatter properties demonstrated excellent performance |
Zhong et al. (2022), Diagnostics 12 | FCM AML diagnostics | Automated gating & AML classification based on multiple ML-based techniques | Rapid and effective technique; integration of other test findings |
Vial et al. (2021), Cancers 13 | MRD in AML | Combined unsupervised FlowSOM & Kaluza software | Powerful tool for MRD, particularly applicable to AML without molecular markers |
Porwit & Bene (2021), Hematology 2 | Plasmacytoid dendritic cell compartment in AL with/without RUNX1 mutation | Unsupervised FCM analysis & clustering | High interpatient variability disclosed by unsupervised analysis |
Reiter et al. (2019), Cytometry Part A 95A | ALL MRD analysis | Multiple Gaussian mixture models (GMM) for automated MRD assessment | Objective & standardized tool for possible use across different laboratories |
Ko et al. (2018), EBioMedicine 37 | MRD in AML and MDS | FCM algorithm for MRD detection (Gaussian mixture model) | High accuracy with short turn-around time; high prognostic significance; ability to integrate with other clinical tests |
Lhermitte et al. (2018), Leukemia 32 | FCM-based diagnosis & classification of AL | Database-guided analysis used for standardized interpretation of the EuroFlow AL orientation tube | Accurate selection of relevant panels for different AL types; computer-supported reproducible classification even without using the full panels |
Quick evaluation of unclear cytopenia
Case 1
Case 2
Case 3
Myelodysplastic neoplasms
Myeloproliferative neoplasms
Myelodyplastic/myeloproliferative neoplasms
Myeloid/lymphoid neoplasms with eosinophilia and tyrosine kinase gene fusions
Mastocytosis
Acute myeloid leukemia
Artificial intelligence and machine learning
Concluding remarks
-
Flow cytometry is a valuable tool for rapid diagnosis, classification, prognosis, and monitoring of hematologic neoplasms.
-
Immunophenotypic profiles can identify underlying genomic alterations and be useful to highlight prognostically relevant differences within subgroups of acute myeloid leukemia.
-
Further standardization and harmonization will become essential for implementing clinical flow cytometry in routine diagnostic evaluations in chronic myeloid neoplasms.
-
Artificial intelligence and machine learning offer an effective way to elaborate and interpret large-scale datasets and help to refine diagnostics.