Use of a hybrid intelligence decision tree to identify mature B-cell neoplasms.

BACKGROUND: Mature B-cell neoplasms are challenging to diagnose due to their heterogeneity and overlapping clinical and biological features. In this study, we present a new workflow strategy that leverages a large amount of flow cytometry data and an artificial intelligence approach to classify these neoplasms. METHODS: By combining mathematical tools, such as classification algorithms and regression tree (CART) models, with biological expertise, we have developed a decision tree that accurately identifies mature B-cell neoplasms. This includes chronic lymphocytic leukemia (CLL), for which cytometry has been extensively used, as well as other non-CLL subtypes. RESULTS: The decision tree is easy to use and proposes a diagnosis and classification of mature B-cell neoplasms to the users. It can identify the majority of CLL cases using just three markers: CD5, CD43, and CD200. CONCLUSION: This approach has the potential to improve the accuracy and efficiency of mature B-cell neoplasm diagnosis.

[Digital morphology analyzers in hematology: a French survey]. / État des lieux des pratiques d'utilisation du microscope automatisé en hématologie en France.

Digital morphology hematology analyzers are becoming more prevalent in laboratories Aims: investigate practices and assess the benefits and limits of digital automated microscopy in hematology. METHODS: questionnaire sent by e-mail in 2018 to French public and private laboratories. RESULTS: out of 118 responses (56 private, 62 public), 117 participants had a CellaVision® microscope, 1 had a West Medica®. Practices were sometimes different, especially in the choice of smears to be digitized or for quality controls (16.1% had internal quality controls, 48.3% external quality controls); 62.1% never used the red blood cell (RBC) characterization tool; the number of cells counted varied from 100 to 400. The study reported a high rate of agreement for these benefits: traceability (95.7%), staff training (94.1%), eye strain (91.4%), risk of error (87.2%), time saving (83.6%). Among the disadvantages, apart from the inadequate search for platelets clumps (93.2%), the agreement rates were often lower: adaptation to digital images (61.2%), difficult assessment of atypical morphologies (49.6%) or RBC morphology (49.6%). CONCLUSION: despite well-established benefits, standardization of practices and technical improvement are still needed.