Related Experiment Video
Updated: Jun 16, 2026
11:27
Prediction of Red Blood Cell Antibody Significance Using the Monocyte-Macrophage Assay
Published on: February 7, 2025
Development of a prediction model for infectious mononucleosis using machine learning algorithms based on blood cell
Yuhong Qin1, Lianjie Zhang1, Hongbin Yu1
1Center of Clinical Laboratory, The Fourth Affiliated Hospital of Soochow University, Suzhou Dushu Lake Hospital, Suzhou, 215213, China.
BMC Infectious Diseases
|June 14, 2026
Summary
This study developed a machine learning model using blood cell counts to accurately diagnose infectious mononucleosis (IM). The Adaptive Boosting Classifier demonstrated high performance, potentially reducing misdiagnosis rates.
Area of Science:
- Hematology
- Infectious Diseases
- Machine Learning
Background:
- Infectious mononucleosis (IM) often presents with nonspecific symptoms, leading to diagnostic challenges and potential complications.
- Current diagnostic methods for IM can have long turnaround times, delaying appropriate patient management.
- There is a need for improved laboratory diagnostic accuracy and efficiency for IM.
Purpose of the Study:
- To develop and optimize a predictive model for infectious mononucleosis (IM) using machine learning algorithms.
- To identify key hematological parameters that contribute to IM diagnosis.
- To enhance the laboratory diagnostic accuracy of IM.
Main Methods:
- Utilized Recursive Feature Elimination (RFE) and cross-validation to select optimal predictive features from patient data.
- Trained eight machine learning algorithms, including the Adaptive Boosting Classifier (AdaBoost), to predict IM.
- Assessed model performance using metrics like AUC, accuracy, sensitivity, specificity, and F1 score, with external validation.
Main Results:
- The AdaBoost model achieved an AUC of 0.928 and an F1 score of 0.800 in the test set.
- Reactive lymphocyte percentage (Reactive lymph%), Lym-Y, and platelet-to-lymphocyte ratio (PLR) were identified as key predictive features.
- The model demonstrated robust performance in an independent validation cohort, with an AUC of 0.923.
Conclusions:
- An IM predictive model integrating the AdaBoost algorithm and three blood cell parameters shows significant diagnostic efficacy.
- This machine learning approach holds promise for reducing missed laboratory diagnoses of IM.
- Further prospective validation is recommended to confirm the clinical utility of the developed model.
