Using Data Mining Models for Differential Diagnosis of Iron Deficiency Anemia and β-thalassemia Minor

Introduction: One of the most common types of anemia is Iron deficiency anemia that its main differential diagnosis is β-thalassemia minor. The rapid and accurate screening of β-thalassemia minor has particular importance for pre-marriage medical counseling and the prevention of the birth of neonates with β-thalassemia major and differentiating it from iron deficiency anemia to avoid unnecessary prescription of iron. The aim of this study was to apply data mining techniques to differentiate iron deficiency anemia from β-thalassemia minor based on CBC test in order to increase the diagnostic speed and to reduce diagnostic costs. Method: The present study was a retrospective study and was performed on 1000 patients referred to Dr. Heidari laboratory of Zahedan city. To conduct research, CRISP-DM standard methodology and support vector machine data mining algorithms, naive Bayes, Bagging, Adaboots and decision tree have been used. WEKA software was used to analyze the data. Results: The results of the evaluations show that Bagging, Decision tree, Adaboots, support vector machine, and naive Bayes algorithms had respectively 95.73%, 95.5%, 94.6%, 80.2% and 76.6% accuracy in differentiating iron deficiency anemia from β-thalassemia minor. Conclusion: In this study, an automatic method based on data mining techniques for differentiation of iron deficiency anemia from β-thalassemia minor is presented. The results of the evaluations show that Bagging algorithm has higher accuracy compared to other data mining algorithms and differential indices. Also, with the help of the decision tree, rules have been extracted that can be used by the physician in timely diagnosis of the two diseases.