Beyond the In-Practice CBC: The Research CBC Parameters-Driven Machine Learning Predictive Modeling for Early...

Area of Science:

• Hematology-oncology
• Artificial intelligence in medicine
• Machine learning for diagnostics

Background:

• Early diagnosis of hematological emergencies, particularly acute leukemias, is crucial for timely treatment.
• Challenges in leukemia diagnosis include symptom overlap, long turnaround times, and the need for specialized expertise.
• Complete blood cell count (CBC) provides morphological and immature cell parameters that may aid in early differentiation.

Purpose of the Study:

• To investigate the potential of AI/ML predictive modeling using CBC-derived cell population data (CPD) for early leukemia differentiation.
• To differentiate between acute and chronic leukemias, and myeloid and lymphoid subtypes at a pre-microscopic level.
• To develop and validate an artificial neural network (ANN) model for leukemia classification.

Main Methods:

• Collected routine and research CBC parameters (CPD) from 1577 patients with hematological neoplasms.
• Utilized statistical tools (heat-map, PCA) to evaluate the predictive capacity of CPD parameters.
• Developed an ANN predictive model driven by CPD parameters to identify disease signatures.

Main Results:

• CPD parameters showed significant deviations across different leukemia types.
• Heat-map and PCA analyses demonstrated effective clustering and segregation of leukemia subtypes (myeloid vs. lymphoid, acute vs. chronic).
• The ANN model achieved high AUC values for differentiating various leukemias (e.g., CML: 0.937, AML: 0.905, ALL: 0.829), with satisfactory training (83.1%) and testing (89.47%) accuracy.

Conclusions:

• Research CBC parameters, analyzed via CPD, hold significant potential for the early differentiation of leukemias.
• CPD-driven ANN modeling offers a novel approach to enhance diagnostic accuracy, acting as a 'disease fingerprint'.
• This AI-powered method can support clinicians in hematology-oncology units for more confident and timely leukemia diagnosis.