Predicting the Risk of Diabetes in Iranian Patients with β-Thalassemia Major / Intermedia Based on Artificial Neural Network

The purpose of this study is to predict catching diabetes in patients having the β-thalassemia. Here, an intelligent system predicts risk of catching diabetes in patients having major thalassemia and intermedia, using multilayer perceptron according to the thalassemia dataset, ZAFAR. In this work, the clinical characteristics of 255 patients, having β-thalassemia, have been studied in two groups of diabetic and non- diabetic patients. Research data includes gender, age, parental family relationship, type of thalassemia, the spleen, gall bladder and liver condition, the age of blood transfusion, blood transfusion intervals, the number of received blood units in each period, the condition of iron overload, the age of start iron chelation therapy, the average of heart T2 rates, the liver T2 and LIC, the number of years having MRI, serum ferritin level, glucose and the number of years the patient was observe. The accuracy of artificial neural network in diagnosing the patients having thalassemia and exposing diabetes would be 80.78% according to the collected dataset. The accuracy rate of the system is 89.48%, using this intelligent system and doing pre-processing. This system has a desirable performance in predicting catching diabetes in patients having β-thalassemia. The best mean square error in this model is 0.07 which results in reducing learning time and increasing the system accuracy. Based on the obtained results from this system, there are two important factors in catching diabetes by the patients having β-thalassemia; first the number of years passed by patient with the high serum ferritin level and second is the glucose level rate in previous years long. The purpose of this study is to predict catching diabetes in patients having the β-thalassemia. Here, an intelligent system predicts risk of catching diabetes in patients having major thalassemia and intermedia, using multilayer perceptron according to the thalassemia dataset, ZAFAR. In this work, the clinical characteristics of 255 patients, having β-thalassemia, have been studied in two groups of diabetic and non- diabetic patients. Research data includes gender, age, parental family relationship, type of thalassemia, the spleen, gall bladder and liver condition, the age of blood transfusion, blood transfusion intervals, the number of received blood units in each period, the condition of iron overload, the age of start iron chelation therapy, the average of heart T2 rates, the liver T2 and LIC, the number of years having MRI, serum ferritin level, glucose and the number of years the patient was observe. The accuracy of artificial neural network in diagnosing the patients having thalassemia and exposing diabetes would be 80.78% according to the collected dataset. The accuracy rate of the system is 89.48%, using this intelligent system and doing pre-processing. This system has a desirable performance in predicting catching diabetes in patients having β-thalassemia. The best mean square error in this model is 0.07 which results in reducing learning time and increasing the system accuracy. Based on the obtained results from this system, there are two important factors in catching diabetes by the patients having β-thalassemia; first the number of years passed by patient with the high serum ferritin level and second is the glucose level rate in previous years long. The purpose of this study is to predict catching diabetes in patients having the β-thalassemia. Here, an intelligent system predicts risk of catching diabetes in patients having major thalassemia and intermedia, using multilayer perceptron according to the thalassemia dataset, ZAFAR. In this work, the clinical characteristics of 255 patients, having β-thalassemia, have been studied in two groups of diabetic and non- diabetic patients. Research data includes gender, age, parental family relationship, type of thalassemia, the spleen, gall bladder and liver condition, the age of blood transfusion, blood transfusion intervals, the number of received blood units in each period, the condition of iron overload, the age of start iron chelation therapy, the average of heart T2 rates, the liver T2 and LIC, the number of years having MRI, serum ferritin level, glucose and the number of years the patient was observe. The accuracy of artificial neural network in diagnosing the patients having thalassemia and exposing diabetes would be 80.78% according to the collected dataset. The accuracy rate of the system is 89.48%, using this intelligent system and doing pre-processing. This system has a desirable performance in predicting catching diabetes in patients having β-thalassemia. The best mean square error in this model is 0.07 which results in reducing learning time and increasing the system accuracy. Based on the obtained results from this system, there are two important factors in catching diabetes by the patients having β-thalassemia; first the number of years passed by patient with the high serum ferritin level and second is the glucose level rate in previous years long.