Journal of Torbat Heydariyeh University of Medical Sciences

---

Background & Aim: Coronary artery disease is among the common diseases in societies. The best method of assessing coronary artery diseases is through angiography. This study aimed at investigating the effect of disease parameters on the diagnosis of coronary artery disease using artificial neural networks.

Methods: This analytic study included a database of 200 non-attributable records. In this research, different neural networks such as MLP, LVQ and BR were used to predict whether the coronary arteries were blocked or not. In addition, the importance of the continuous risk factors of coronary artery disease was studied.

Results: The most important criteria of the diagnosis systems are the specificity and sensitivity indicators. In this study, these two indicators were calculated in the test. The best accuracy was observed in MLP, with a back-error propagation of 88%. It was also observed that the removal of discrete parameters positively affects neural network convergence speed so that the prediction accuracy could reach 85%.

Conclusion: Angiography is a high-cost invasive procedure with risk factors such as death, stroke and heart attack. Therefore, noninvasive methods should be applied in order to minimize   error and maximize reliability to predict the disease. Using data mining methods can decrease the complications of the disease.

---

Background & Aim: A main problem in diabetes is its timely and accurate diagnosis. This study aimed at diagnosing diabetes using data mining methods.
Methods: The present study is an analytical investigation including 768 individuals with 8 attributes. Artificial neural networks and fuzzy neural networks were used to diagnose the diabetes. To achieve a real accuracy, the Kfold method was used to divide samples into training and test groups.
Results: The mean square errors in multilayer perceptron network (MLP), learning vector quantization and Nero fuzzy networks were 98.6%, 98.2% and 99.6%, respectively.
Conclusion: According to the results of this study, , data mining method can be effective in diagnosing diabetes.  In this regard, both used methods are useful; however, higher precision was obtained following the use of Neuro-Fuzzy approach.

---

Background & Aim: Coronary artery disease is one of the most common diseases in different societies. Coronary angiography is established as one of the best methods for diagnosis of this disease. Angiography is an invasive and costly method. Furthermore, it is associated with risks such as death, heart attack, and stroke. Thus, this study introduces a neuro-fuzzy-based method which can help the physicians in prediction of patient’s coronary artery condition.
Methods: This is an analytical study carried on 200 patients of Cardiovascular Center in Torbat Heydarieh. Patient records include 13 risk factors and are non-attributable. In this work, models are presented based on data mining methods for the diagnosis of coronary artery disease Furthermore, artificial neural network and neuro-fuzzy method were used for modeling the diagnosis of coronary artery disease.
Results: The mean square error (MSE) of prediction for artificial neural network and neuro-fuzzy method were p=0.2574 and p=0.0007, respectively.
Conclusion: Since angiography is invasive and associated with various risks, we suggest the use of non-invasive methods with low error and high reliability. New data mining strategies can be effective in reducing the mentioned complications.

---

Background & Aim: Detecting the abnormal performance of diabetes and subsequently getting proper treatment can reduce the mortality associated with the disease. Also, timely diagnosis will result in irreversible complications for the patient. The aim of this study was to determine the status of diabetes mellitus using data mining techniques.
Methods: This is an analytical study and its database contains 254 independent records based on 13 characteristics. Data is collected by a researcher from one of the specialized diabetes centers in Mashhad.
Results: After preprocessing of the obtained data, different methods of pattern recognition were applied. Using multilevel MLP neural networks, LVQ neural networks, SVM support vector and Kmeans clustering method, the mean square error (RMSE) was calculated. The correctness of each learner's performance is 94%, 92%, 96%, and 93%, respectively.
Conclusion: Reducing the diagnosis of diabetes is one of the goals of the researchers. Using data mining techniques can help to reduce this error. In this study, among different protocols used for pattern recognition, SMV method displayed a significantly better performance.

---

Background & Aim: Autism spectrum disorder is one of the psychological disorders in children. Timely and accurate diagnosis of this disorder is of great importance in providing appropriate care and treatment for children. The main objective of this research is to emphasize the significance of features related to autism and their diagnosis using an intelligent model, as some of these features have higher priority.
Methods: For this purpose, the principal component analysis (PCA) method was employed to prioritize the features, and after extracting the optimal features, automatic disease diagnosis was performed using artificial neural networks.
Results: The data used in this study were collected from the Kaggle dataset, including 1054 individuals, out of which 728 were diagnosed with autism and 326 were healthy. The results of this study indicate that the gradual elimination of features and reduction from 18 to 12 features can lead to achieving the same accuracy in diagnosing the autism spectrum using artificial neural networks.
Conclusion: Reducing the number of features in artificial intelligence models for autism diagnosis not only improves and optimizes the diagnostic process but also helps in reducing parental stress and preserving their privacy due to the reduced number of questions. Ultimately, this leads to the generation of models with better performance and interpretability.