Abnormal characteristic "round bottom flask" shape volume-based scattergram as a trigger to suspect persistent...

Area of Science:

• Hematology
• Computational Biology
• Diagnostic Technology

Background:

• Diagnosing persistent polyclonal B-cell lymphocytosis (PPBL) is challenging due to overlapping features with splenic marginal zone lymphomas (SMZL).
• Current diagnostic methods lack specific markers for PPBL, necessitating novel approaches.

Purpose of the Study:

• To develop a machine learning (ML) model for improved detection and diagnosis of PPBL.
• To leverage data from the DxH 800 analyzer, including cell population data (CPD) and scattergrams, for classification.

Main Methods:

• A total of 211 samples (101 controls, 110 patients with PPBL or SMZL) were analyzed.
• Data included full blood count, CPD, scattergrams, flags, and CellaVision differentials.
• A machine learning model was constructed using these parameters for classification.

Main Results:

• PPBL and SMZL shared elevated lymphoid counts, atypical lymphoid flags, and specific CPD values.
• A unique "round-bottom-flask" scattergram pattern was identified for PPBL, also observed in SMZL cases.
• The ML model achieved 93.4% classification accuracy, correctly identifying all pathological cases with low misclassification rates.

Conclusions:

• The developed ML model offers a novel, unbiased tool for laboratory use.
• This approach can significantly aid in the detection and diagnosis of PPBL.
• The model's high accuracy demonstrates its potential for widespread clinical application.