Yazar "Tajally, AmirReza" seçeneğine göre listele
Listeleniyor 1 - 3 / 3
Sayfa Başına Sonuç
Sıralama seçenekleri
Öğe A framework for robust glaucoma detection: A confidence-aware deep uncertainty quantification approach with a comprehensive assessment for enhanced clinical decision-making(Elsevier ltd, 2025) Zarean, Javad; Tajally, AmirReza; Tavakkoli-Moghaddam, Reza; Sajadi, Seyed Mojtaba; Wassan, NiazGlaucoma poses a significant threat to public health worldwide, as it can result in irreversible vision loss. Timely identification is vital for halting the progression of visual field deterioration. In recent years, deep neural networks (DNNs) have become increasingly popular in medical imaging due to their ability to identify patterns. As a result, this study introduces a new computer-aided diagnosis (CAD) system based on deep learning (DL) algorithms for glaucoma detection that extracts meaningful features from retinal fundus images (RFIs) and employs uncertainty quantification (UQ) models, including Monte Carlo dropout (MCD), ensemble Bayesian, and ensemble Monte Carlo dropout (EMCD), to generate both point estimates and confidence values for the outputs, thereby capturing the uncertainty associated with the classifications. The proposed framework is validated using well-known clinical datasets, and the reliability of the outputs is evaluated using comprehensive performance metrics such as expected calibration error (ECE), entropy analysis, and a multi-criteria UQ assessment. Experimental results demonstrate the superiority of the ensemble model, with uncertainty accuracies registering at 97.64%, 97.26%, and 98.97% for the "ACRIMA", "RIM-ONE-DL", and "ORIGA" datasets, respectively. Moreover, the proposed algorithms can alert users to the majority of erroneous diagnoses by assigning uncertainty labels, providing valuable insights for clinicians in glaucoma detection. Such tools can assist healthcare professionals in reducing the probability of misdiagnosis and ensuring that patients receive timely and appropriate treatment.Öğe A hybrid machine learning model based on ensemble methods for devices fault prediction in the wood industry(Pergamon-Elsevier Science Ltd, 2024) Dahesh, Arezoo; Tavakkoli-Moghaddam, Reza; Wassan, Niaz; Tajally, AmirReza; Daneshi, Zahra; Erfani-Jazi, AsemanIn manufacturing industries, including the wood industry, devices, and equipment are considered the basic elements and the main capital for production. That is why managers are trying to maintain and use these devices and equipment optimally. On the other hand, repurchasing device parts or repairing equipment in case of major damage can cause more damage than planned costs. Therefore, a model that can determine the fault class based on the signs seen in the equipment would prevent major damage to the device and save on repair costs. In this regard, using the registered features for equipment and with the help of machine learning algorithms, a model can be created that can classify devices in the appropriate class based on their observed features. The present study uses nine machine learning algorithms to make this model, trains each model on three sets of selected features, and finally compares them. It is worth mentioning that after evaluating the models, based on the features selected from the embedded techniques, permutation feature importance methods, and genetic algorithm, the best models are considered as categorical boosting with the training and testing accuracy of 0.895 and 0.909, random forest with the training and testing accuracy of 0.905 and 0.893, and extreme gradient boosting with the training and testing accuracy of 0.884 and 0.885.Öğe Robust electroencephalogram-based biometric identification against GAN-generated artificial signals using a novel end-to-end attention-based CNN-LSTM neural network(Springer, 2025) Zarean, Javad; Tajally, AmirReza; Tavakkoli-Moghaddam, Reza; Kia, RezaElectroencephalogram (EEG) signals, which exhibit dynamic properties and discriminative information among individuals, have recently been employed to develop human biometric identification and authentication systems. Despite the increasing interest in EEG-based human identification, the state-of-the-art still needs high-accuracy and easy-to-use systems in real-life applications. To improve the accuracy, robustness, and user-friendliness of EEG-based human identification systems, this paper presents a novel attention-based convolutional-long short-term memory network for EEG-based human biometric identification (ABCL-EHBI), which is robust against artificial EEG signals generated by generative adversarial networks (GANs). The proposed system uses an attention mechanism along with convolutional neural networks (CNNs) and long short-term memories (LSTMs) layers, leading to more effective exploitation of the raw EEG signals' spatial and temporal discriminative characteristics compared to a simple CNN-LSTM (CL) system. The system was evaluated and validated using the PhysioNet motor imagery dataset, which incorporates EEG signals of 109 individuals performing six various tasks. Experimental results show that the proposed approach achieves F1-Score accuracy of 99.65, 99.64, and 99.55 under the condition of using 64, 14, and only 9 EEG channels, respectively, which is better than the performance of EEG-based human identification in the previous studies. The fact that the proposed approach receives raw EEG signals as input without any need for feature extraction (end-to-end), shows high accuracy when using a small number of EEG channels and yields high accuracy against artificial EEG signals, making it reliable and easy to deploy in real-life applications.
