A novel method for screening malignant hematological diseases by constructing an optimal machine learning model based...

Area of Science:

• Hematology
• Medical Informatics
• Artificial Intelligence

Background:

• Screening for malignant hematological diseases is crucial for timely diagnosis and treatment.
• Routine blood cell parameters offer a valuable basis for developing screening models.
• Early detection significantly improves patient outcomes in hematological malignancies.

Purpose of the Study:

• To construct an optimal screening model for malignant hematological diseases.
• To leverage routine blood cell parameters for enhanced diagnostic capabilities.
• To develop a reliable tool for identifying hematological malignancies.

Main Methods:

• Collected venous blood samples from 1751 patients across 10 tertiary hospitals in China.
• Utilized a training set (1223 cases) and a validation set (528 cases).
• Developed eight machine learning models, including an artificial neural network (ANN), using 26 blood cell parameters and clinical data.

Main Results:

• The artificial neural network (ANN) model demonstrated optimal performance in the validation set.
• The ANN model achieved an AUC of 0.906, accuracy of 0.857, sensitivity of 0.832, and specificity of 0.884.
• This indicates high discrimination, calibration, and clinical detection performance for identifying malignant hematological diseases.

Conclusions:

• The developed ANN model is suitable for screening malignant hematological diseases.
• This model is particularly beneficial for primary hospitals lacking comprehensive diagnostic facilities.
• Early screening using the ANN model can expedite diagnosis and treatment for patients.