Related Experiment Video
Updated: May 17, 2025
05:35
Author Spotlight: Developing a Point-of-Care Hemoglobin Estimation Method for Anemia Management
Published on: January 19, 2024
688
Machine learning approach for differentiating iron deficiency anemia and thalassemia using random forest and gradient
Wanicha Tepakhan1,2, Wisarut Srisintorn3, Tipparat Penglong1
1Department of Pathology, Faculty of Medicine, Prince of Songkla University, Hat Yai, Songkhla, Thailand.
Scientific Reports
|May 15, 2025
Summary
Machine learning models effectively distinguish iron deficiency anemia (IDA) from thalassemia (Thal). Gradient boosting and random forest algorithms show high accuracy in diagnosing these conditions, aiding in endemic regions.
Area of Science:
- Hematology
- Computational Biology
- Medical Diagnostics
Background:
- Red blood cell indices are used to differentiate iron deficiency anemia (IDA) and thalassemia (Thal).
- Existing formulas show variable diagnostic efficiency.
- A need exists for improved diagnostic tools for IDA and Thal.
Purpose of the Study:
- To develop and evaluate machine learning algorithms for discriminating between IDA and Thal.
- To compare the performance of random forest (RF) and gradient boosting (GB) algorithms.
Main Methods:
- Utilized complete blood count data from 1143 patients with anemia and low mean corpuscular volume.
- Data included 382 IDA, 635 Thal, and 126 mixed IDA and Thal cases.
- Employed RF and GB algorithms, with data split into 80:20 training and testing sets.
Main Results:
- Both RF and GB models demonstrated strong diagnostic performance in predicting IDA and Thal.
- In binary outcome prediction (testing set), GB and RF achieved 90.7% accuracy and 0.953 AUC-ROC.
- Performance decreased when including patients with both IDA and Thal, with accuracies of 80.4% (GB) and 82.2% (RF).
Conclusions:
- A machine learning approach, particularly using the GB algorithm, was successfully developed.
- This tool shows potential utility for diagnosing IDA and Thal in endemic areas.
- Further refinement may be needed for cases involving co-existing IDA and Thal.
