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

Background: Early diagnosis plays a crucial role in improving the survival rate of acute leukemia (AL) patients. This study aims to develop a warning model for the detection of acute leukemia (AL) using complete blood count (CBC) and cell population data (CPD), which could aid in clinical diagnosis.

Methods: In this study, CBC and CPD were utilized to develop a warning model for assisting clinical diagnosis of AL. Clinical characteristics and peripheral blood data were retrospectively collected from 262 AL patients and 280 non-AL patients at the Second Hospital of Lanzhou University; they were randomly divided into a training set and a test set in a ratio of 7:3. The training set was used to establish support vector machine (SVM), random forest (RF), and logistic regression (LR) models for AL. The validation set consisted of 357 cases (97 AL, 260 non-AL) collected from the General Hospital of Ningxia Medical University to verify the warning efficacy of the optimal model in conjunction with the test set.

Results: The comparative analysis revealed that the SVM model outperformed the RF and LR models in terms of diagnostic accuracy. In the training set, the accuracy was 92.93%; the area under the ROC curve (AUC) and 95% confidence interval (95% CI) were 0.981 (0.970, 0.992). For the test set, the accuracy was 89.66%; the AUC and 95% CI were 0.959 (0.931, 0.988). As for the validation set, the accuracy was 76.34%; the AUC and 95% CI were 0.841 (0.789, 0.893). Additionally, the calibration curve and decision curve analysis (DCA) demonstrated that the SVM model exhibited satisfactory effectiveness and feasibility.

Conclusion: The SVM model shows significant potential as a clinical screening tool for AL.