► Click here to download the original poster (PDF)
Julie BLANCHI1, Sebastien RAIMBAULT2, Benoit RUCHETON1, Marion EVEILLARD3, Françoise DURRIEU1
1Institut Bergonié, Bordeaux, France; 2HORIBA ABX SAS, Grabels, France; 3University Hospital, Nantes, France
Myelodysplastic neoplasms (MDS) are often challenging to diagnose due to the wide range of alternative causes of cytopenia, particularly inflammatory or immune-mediated disorders. Diagnosis requires bone marrow (BM) evaluation along with advanced techniques such as flow cytometry (FCM), karyotyping, and molecular analysis. This study investigates the diagnostic performance of the Yumizen H2500 hematology analyzer in identifying MDS in patients presenting with cytopenia. The objective is to evaluate the potential utility of Full Blood Count with differential (FBC-DIF), reticulocyte parameters (DIR), and optical platelet measurement (PLT-Ox) channels in differentiating MDS-associated cytopenia from other etiologies, thereby potentially reducing the need for invasive BM examinations.
The study involved 252 patients with at least one cytopenia (White blood cell count <4x10^9/L, Hemoglobin <12 g/dL, Platelet count<150x10^9/L) who underwent bone marrow analysis. Blood samples were analyzed using the Yumizen H2500 analyzer (HORIBA, France) in different modes (DIF, DIR or PLT-Ox). Additionally, blood immunophenotyping was conducted on a Dx Flex cytometer (Beckman-Coulter, USA). Machine-learning models were applied to identify the most relevant variables, with stratified 20 fold cross-validation performed using Orange-ML v3.38 software (Fig.1).
MDS diagnosis (MDS pos) was confirmed in 101 patients (40,08%) by BM cytology, immunophenotyping, cytogenetic and molecular analysis. Flow cytometry (FCM) blood analysis corroborated the accuracy of the Yumizen H2500 differential and its ability to detect MDS-related abnormalities, particularly in granulocyte and monocyte scatter patterns (Fig. 2). The Gradient Boosting algorithm was the most effective for detecting MDS for DIF and DIR. No significant contribution was observed from the Reticulocyte parameters. The Optical mode selected five relevant parameters, with the Support Vector Machine (SVM) model performing best for this subset. A scoring system (DIF-score, DIR-score, PLT-Ox score) was developed with strong positive predictive value (PPV) and negative predictive value (NPV) for MDS detection. Results are summarized in Table 1.
Figure 1: MDS flag analysis flowchart
Figure 2: Comparison of granulocyte and monocyte scatter patterns between the Yumizen H2500 (left panel) and FCM (right panel)
(A) and (B) illustrate a reactive cytopenia sample; (C) and (D) illustrate a MDS sample with granulocyte abnormal pattern.
FCM: Blue (lymphocytes); Green (granulocytes); Purple (monocytes).
Yumizen H2500: Blue (lymphocytes), Green (neutrophils), Purple (monocytes), Orange (eosinophils)
| MDS-score | |||||
|---|---|---|---|---|---|
| Channel | AUC | Sensitivity | Specificity | PPV | NPV |
| DIF-score | 0.841 | 65.0% | 90.6% | 82.5% | 79.1% |
| DIR-score | 0.856 | 74.5% | 89.3% | 83.7% | 82.7% |
| PLT-Ox score | 0.800 | 61.8% | 94.9% | 91.3% | 74.0% |
Table 1: Statistical performance metrics of MDS score for the different models used in the study.
AUC: Area Under the Curve, PPV: Positive Predictive Value, NPV: Negative Predictive Value
This study shows the potential effectiveness of the HORIBA Yumizen H2500 MDS score in identifying MDS. Proper use of cell population data could be clinically beneficial in routine practice. Incorporating the MDS flag into FBC reports could optimize Turn Around Time (TAT) and reduce unnecessary and expensive bone marrow investigations by providing qualitative NPV results. Future work will focus on validating the MDS score with a separate cohort.
Vous avez des questions ou des demandes ? Utilisez ce formulaire pour contacter nos spécialistes.
