Cell Population Data-Driven Acute Promyelocytic Leukemia Flagging Through Artificial Neural Network Predictive...

Area of Science:

• Hematology
• Computational Biology
• Oncology

Background:

• Acute promyelocytic leukemia (APML) requires timely treatment for better patient outcomes.
• Early detection of APML is crucial for initiating prompt therapeutic interventions.
• Traditional diagnostic methods can be time-consuming, necessitating faster screening tools.

Purpose of the Study:

• To develop and validate an artificial neural network (ANN) model for the early detection of APML.
• To identify key complete blood cell count (CBC) parameters and cell population data (CPD) for APML prediction.
• To enhance the diagnostic accuracy of routine hematological tests for APML.

Main Methods:

• Collected CBC and CPD from 1067 hematological neoplasm patients.
• Utilized statistical analysis and principal component analysis (PCA) for parameter evaluation.
• Developed an ANN predictive model using selected CBC parameters to differentiate APML cases.

Main Results:

• Identified a characteristic triad for APML: low platelet count (PLT), decreased/normal immature platelet fraction (IPF), and high neutrophil DNA/RNA content (NE-SFL).
• PCA confirmed significant variance in PLT, IPF, and NE-SFL for APML.
• The ANN model achieved a high area under the curve (AUC) of 0.894 with a low false prediction rate (2.3%) for APML classification.

Conclusions:

• The combination of PLT, IPF, and NE-SFL parameters shows potential for early APML flagging.
• CBC item-driven ANN modeling offers a novel and effective approach to improve diagnostic prediction in hematology.
• This method can assist clinicians in confidently identifying APML cases through typical parameter trends.