Friday, March 29, 2024
HomeAmerican Journal of Clinical Pathologyindex/list_12094_1A Machine Learning Approach to the Classification of Acute Leukemias and Distinction...

A Machine Learning Approach to the Classification of Acute Leukemias and Distinction From Nonneoplastic Cytopenias Using Flow Cytometry Data

Abstract and Introduction

Abstract

Objectives: Flow cytometry (FC) is critical for the diagnosis and monitoring of hematologic malignancies. Machine learning (ML) methods rapidly classify multidimensional data and should dramatically improve the efficiency of FC data analysis. We aimed to build a model to classify acute leukemias, including acute promyelocytic leukemia (APL), and distinguish them from nonneoplastic cytopenias. We also sought to illustrate a method to identify key FC parameters that contribute to the model’s performance.

Methods: Using data from 531 patients who underwent evaluation for cytopenias and/or acute leukemia, we developed an ML model to rapidly distinguish among APL, acute myeloid leukemia/not APL, acute lymphoblastic leukemia, and nonneoplastic cytopenias. Unsupervised learning using gaussian mixture model and Fisher kernel methods were applied to FC listmode data, followed by supervised support vector machine classification.

Results: High accuracy (ACC, 94.2%; area under the curve [AUC], 99.5%) was achieved based on the 37-parameter FC panel. Using only 3 parameters, however, yielded similar performance (ACC, 91.7%; AUC, 98.3%) and highlighted the significant contribution of light scatter properties.

Conclusions: Our findings underscore the potential for ML to automatically identify and prioritize FC specimens that have critical results, including APL and other acute leukemias.

Introduction

Flow cytometry (FC) immunophenotypic analysis is a critical component of testing to establish precise diagnoses for hematolymphoid neoplasms and monitor therapeutic response.[1–3] Computational methods to evaluate cytometry data have been evolving for exploratory and discovery research,[4,5] but with the exception of tools that the EuroFlow Consortium has developed,[6,7] clinical software has primarily provided a user interface an analyst can use to manually inspect and manipulate data displayed on 2-dimensional plots through a complex, sequential gating process.[8] Because this approach is labor intensive, heavily dependent on specialized expertise, and difficult to standardize, data analysis has become a rate-limiting factor for providing the FC interpretations needed for patient care. A solution for this bottleneck would increase laboratory efficiency and permit more rapid diagnoses of acute leukemias and other hematologic malignancies.

Artificial intelligence (AI), including machine learning (ML), has the potential to substantially assist physicians caring for patients with hematolymphoid diseases with interpreting and using complex data for diagnosis, risk stratification, and response prediction.[9,10] ML models have demonstrated human-level performance using FC data to classify B-cell neoplasms[11,12] and detect residual leukemia.[13,14] Our ML approach to rapidly classifying FC data (~7 seconds) predicted residual acute myeloid leukemia (AML) and myelodysplastic syndrome (MDS) with promising accuracy (84.6%-92.4%) and was associated with survival.[14] Whether a similar approach could be used to distinguish leukemic from nonneoplastic bone marrow samples and to rapidly distinguish acute promyelocytic leukemia (APL) from AML and acute lymphoblastic leukemia (ALL) was uncertain.

To further investigate the application of ML approaches using clinical FC data, we aimed to build a model to classify acute leukemias, including APL, and distinguish them from nonneoplastic cytopenias. We also sought to illustrate a method to identify key FC parameters that contribute to the model’s performance. Our findings highlight the potential for AI to support clinical FC laboratories to efficiently detect and classify hematolymphoid neoplasms.

RELATED ARTICLES
- Advertisment -

Most Popular