Azam Orooji , Mostafa Langarizadeh , Maryam Aghazadeh, Mehran Kamkarhaghighi, Marjan Ghazisaiedi , Fateme Moghbeli,
Volume 12, Issue 4 (11-2018)
Abstract
Background and Aim: Artificial intelligence is a branch of computer science that has the ability of analyzing complex medical data. Using artificial intelligence is common in diagnosing, treating and taking care of patients. Warfarin is one of the most commonly prescribed oral anticoagulants. Determining the exact dose of warfarin needed for patients is one of the major challenges in the health system, which has attracted the attention of researchers. The purpose of this study was to determine the exact dose of warfarin needed for patients with artificial heart valves using artificial neural networks (ANN).
Materials and Methods: A total of 9 multi-layer perceptron ANNs with different structures were constructed and evaluated based on a dataset including 846 patients who had referred to the PT clinic in Tehran Heart Center in the second half of the year 2013. Finally, the best structure of ANN for warfarin dose was investigated. All simulations including data preprocessing and neural network designing were done in MATLAB environment.
Results: The effectiveness of ANNs was evaluated in terms of classification performance using 10-fold cross-validation procedure and the results showed that the best model was a network that had 7 neurons in its hidden layer with an average absolute error of 0.1, turbulence rate of 0.33, and regression of 0.87.
Conclusion: The achieved results reveal that ANNs are able to predict warfarin dose in Iranian patients with an artificial heart valve. Although no system can be guaranteed to achieve 100% accuracy, they can be effective in reducing medical errors.
Niloofar Mohammadzadeh, Ziba Mosayebi, Hamid Beigy, Mohammad Shojaeinia,
Volume 14, Issue 6 (1-2021)
Abstract
Background and Aim: Sepsis is the most important disease in the first 28 days of life and one of the main causes of infant mortality in the intensive care unit. Its definitive diagnosis is possible by performing blood culture. Neonatal sepsis can be a clinical sign of nosocomial infections that are often resistant to antibiotics. Therefore, the purpose of this study was to create and evaluate a hospital sepsis prediction model and present its results to health care providers.
Materials and Methods: In this descriptive-applied study, the research population includes neonates admitted to the intensive care unit of Valiasr Hospital in Tehran and the research sample is the data of 4196 neonates admitted to this ward from 2016 to August, 2020. The initial features for creating a predictive model of sepsis were prepared by examining the relevant information sources and under the supervision of professors and officials of Valiasr Hospital's mother and fetus research center and its validity was confirmed by 5 neonatal professors of this hospital. In this research, machine learning algorithms have been used to create a sepsis prediction model.
Results: Accuracy and AUROC(area under the ROC curve) parameters were used to evaluate the generated models. The highest values of Accuracy and AUROC are related to Adaptive Boosting and random forest algorithms, respectively.
Conclusion: Learning curves show that using different training examples and more complex selection of combination features improves the performance of the models. Further research is needed to evaluate the clinical effectiveness of machine learning models in a trial.
Marsa Gholamzadeh, Seyed Mohammad Ayyoubzadeh, Hoda Zahedi, Sharareh Rostam Niakan Kalhori,
Volume 15, Issue 3 (8-2021)
Abstract
Background and Aim: Due to the important role of radiological images for identifying patients with COVID-19, creating a model based on deep learning methods was the main objective of this study.
Materials and Methods: 15,153 available chest images of normal, COVID-19, and pneumonia individuals which were in the Kaggle data repository was used as dataset of this research. Data preprocessing including normalizing images, integrating images and labeling into three categories, train, test and validation was performed. By Python language in the fastAI library based on convolution technique (CNN) and four architectures (ResNet, VGG MobileNet, AlexNet), nine models through transitional learning method were trained to recognize patients from healthy persons. Finally, the performance of these models was evaluated with indicators such as accuracy, sensitivity and specificity, and F-Measure.
Results: Of the nine generated models, the ResNet101 model has the highest ability to distinguish COVID-19 cases from other cases with 95.29% sensitivity. Other applied models showed more than 96% accuracy in correctly diagnosis of various cases in test phase. Finally, the ResNet101 model was able to demonstrate 98.4% accuracy in distinguishing between healthy and infected cases.
Conclusion: The obtained accuracy showed the accurate performance of developed model in detecting COVID-19 cases. Therefore, by implementing an application based on the developed model, physicians can be helped in accurate and early diagnosis of cases. an application based on the developed model, physicians can be helped in accurate and early diagnosis of infected cases.
Miss Fariba Moalem Borazjani, Azita Yazdani, Reza Safdari, Seyed Mansoor Gatmiri,
Volume 17, Issue 6 (2-2024)
Abstract
Background and Aim: Kidney failure is a common and increasing problem in Iran and worldwide. Kidney transplantation is recognized as a preferred treatment method for patients with end-stage renal disease (ESRD). Machine learning, as one of the most valuable branches of artificial intelligence in the field of predicting patient outcomes or predicting various conditions in patients, has significant applications. The purpose of this research was to predict kidney transplant outcomes in patients using machine learning.
Materials and Methods: Since CRISP is one of the strongest methodologies for implementing data mining projects, it was chosen as the working method. In order to identify the factors affecting the prediction of kidney transplant outcomes, a researcher-created checklist was sent to some of nephrologists nationwide to determine the importance of each factor. The results were analyzed and examined. Then, using Python language and different algorithms such as random forest, SVM, KNN, deep learning, and XGBoost the data was modeled.
Results: The final model was multilabel, capable of predicting various kidney transplant outcomes, including rejection probability, diabetic reactions, malignant reactions, and patient rehospitalization. After modeling the input data features, the model was able to predict the four kidney transplant outcomes such as rejection, diabetes, malignancy and readmission with an error rate of less than 0.01.
Conclusion: The high level of accuracy and precision of the random forest model demonstrates its strong predictive power for forecasting kidney transplant outcomes. In this study, the most influential factors contributing to patient susceptibility to the mentioned outcomes were identified. Using this machine learning-based system, it is possible to predict the probability of these outcomes occurring for new cases.
