Tehran University Medical Journal

Background: Gestational diabetes mellitus (GDM) is one of the most common metabolic disorders in pregnancy, which is associated with serious complications. In the event of early diagnosis of this disease, some of the maternal and fetal complications can be prevented. The aim of this study was to early predict gestational diabetes mellitus by two statistical models including artificial neural network (ANN) and decision tree and also comparing these models in the diagnosis of GDM.
Methods: In this modeling study, among the cases of pregnant women who were monitored by health care centers of Kermanshah City, Iran, from 2010 to 2012, four hundred cases were selected, therefore the information in these cases was analyzed in this study. Demographic information, mother's maternal pregnancy rating, having diabetes at the beginning of pregnancy, fertility parameters and biochemical test results of mothers was collected from their records. Perceptron ANN and decision tree with CART algorithm models were fitted to the data and those performances were compared. According to the accuracy, sensitivity, specificity criteria and surface under the receiver operating characteristic (ROC) curve (AUC), the superior model was introduced.
Results: Following the fitting of an artificial neural network and decision tree models to data set, the following results were obtained. The accuracy, sensitivity, specificity and area under the ROC curve were calculated for both models. All of these values were more in the neural network model than the decision tree model. The accuracy criterion for these models was 0.83, 0.77, the sensitivity 0.62, 0.56 and specificity 0.95, 0.87, respectively. The surface under the ROC curve in ANN model was significantly higher than decision tree (0.79, 0.74, P=0.03).
Conclusion: In predicting and categorizing the presence and absence of gestational diabetes mellitus, the artificial neural network model had a higher accuracy, sensitivity, specificity, and surface under the receiver operating characteristic curve than the decision tree model. It can be concluded that the perceptron artificial neural network model has better predictions and closer to reality than the decision tree model.

With the rapid expansion of artificial intelligence across clinical disciplines, a variety of artificial neural networks (ANNs) have become indispensable tools for endowing computer systems with advanced analytical power. Dentistry, as an information‑rich branch of medicine, routinely generates and must interpret large, complex datasets from imaging and diagnostic records. Consequently, researchers have increasingly directed their attention toward intelligent, automated techniques for analyzing dental data. This study therefore surveys and synthesizes the methods that have been applied to the intelligent and automated analysis of such data, highlighting the prevailing trends in current literature.The majority of the examined investigations relied on panoramic radiographic images of the teeth orthopantomograms (OPG) as their primary source material. Three overarching technical objectives repeatedly emerged: first, tooth diagnosis, meaning the reliable separation and identification of each individual tooth from its neighbors; second, sample segmentation, that is, the piece‑by‑piece analysis of visual information within the image; and third, semantic segmentation, namely, the contextual interpretation of information extracted from the radiograph. Depending upon which of these objectives was pursued, researchers selected different neural‑network architectures and configurations. Across the reviewed corpus, input images were typically subjected to preprocessing steps such as normalization, noise reduction, and contrast enhancement before being supplied to a neural network for training, thereby preparing the data for subsequent machine interpretation. In several instances, the raw output produced by the neural network underwent additional post‑processing, a stage designed to refine the preliminary results and enhance overall accuracy. The comparative analysis presented here concentrates on how effectively the various neural‑network models fulfilled the three technical objectives described above. The surveyed articles reveal two dominant analytical approaches. In the intelligent problem‑solving paradigm, convolutional neural networks (CNNs) overwhelmingly predominate. Conversely, in the automated paradigm, investigators favor classical, non‑learning algorithmic techniques. Work employing ANNs consistently emphasizes image comprehension, segmentation, feature extraction, feature classification, network modeling, and careful variable tuning to promote effective learning that aligns with each study’s stated objectives.