Acute Leukemia Warning Model Combined CBC and CPD Data Based on Machine Learning.

Area of Science:

• Hematology
• Medical Informatics
• Machine Learning in Healthcare

Background:

• Early diagnosis of acute leukemia (AL) is critical for improving patient survival rates.
• Complete blood count (CBC) and cell population data (CPD) are valuable for disease detection.
• Developing effective warning models can assist clinicians in diagnosing AL.

Purpose of the Study:

• To develop and evaluate a warning model for acute leukemia (AL) detection.
• To utilize complete blood count (CBC) and cell population data (CPD) for AL screening.
• To compare the performance of different machine learning models in AL diagnosis.

Main Methods:

• Retrospective collection of CBC and CPD data from AL and non-AL patients.
• Development of Support Vector Machine (SVM), Random Forest (RF), and Logistic Regression (LR) models.
• Validation of the optimal model using independent test and validation datasets.

Main Results:

• The SVM model demonstrated superior diagnostic accuracy compared to RF and LR models.
• SVM model achieved high accuracy (92.93% training, 89.66% test) and AUC (0.981 training, 0.959 test).
• The SVM model showed satisfactory effectiveness and feasibility in external validation (76.34% accuracy, 0.841 AUC).

Conclusions:

• The Support Vector Machine (SVM) model shows significant potential as a clinical screening tool for acute leukemia (AL).
• The developed model effectively utilizes CBC and CPD for AL detection.
• The SVM model's performance supports its feasibility for aiding clinical diagnosis.