ThalPred

ThalPred discriminates thalassemia trait (TT) from iron deficiency anemia (IDA) using machine learning on red blood cell (RBC) indices to support accurate differentiation of hypochromic microcytic anemia (HMA).

Key Features:

• Data-driven dataset: Retrospective laboratory data from 186 Thai adults (146 TT, 40 IDA) were used for model training and validation.
• Machine learning algorithms: Five algorithms—k-nearest neighbor (k-NN), decision tree, random forest (RF), artificial neural network (ANN), and support vector machine (SVM)—were trained as discriminant models.
• Comparative evaluation: Models were evaluated against thirteen existing discriminant formulas and indices for distinguishing IDA and TT.
• Optimal model performance: The SVM model achieved external accuracy of 95.59%, Matthews correlation coefficient (MCC) of 0.87, and area under the curve (AUC) of 0.98.
• Interpretable rules: Rules derived from the RF model describe combinations of RBC indices that differentiate IDA from TT.

Scientific Applications:

• Differential diagnosis of HMA: Supporting screening and discrimination between TT and IDA using RBC indices.
• Clinical decision support: Informing treatment stratification and management decisions for patients with microcytic anemia.
• Method benchmarking: Providing a comparative framework for assessing discriminant formulas and machine learning classifiers in hematology research.

Methodology:

Retrospective laboratory data from 186 Thai adults (146 TT, 40 IDA) were used to train k-NN, decision tree, RF, ANN, and SVM models, compare them to thirteen discriminant formulas/indices, select the SVM model based on external accuracy (95.59%), MCC (0.87), and AUC (0.98), and extract interpretable rules from the RF model.