Objective: To evaluate the performance of vision transformer (ViT)-based deep learning models in the classification of open apex on panoramic radiographs (orthopantomograms (OPGs)) and compare their diagnostic accuracy with conventional convolutional neural network (CNN) architectures. Materials and Methods: OPGs were retrospectively collected and labeled by two observers based on apex closure status. Two ViT models (Base Patch16 and Patch32) and three CNN models (ResNet50, VGG19, and EfficientNetB0) were evaluated using eight classifiers (support vector machine (SVM), random forest (RF), XGBoost, logistic regression (LR), K-nearest neighbors (KNN), naïve Bayes (NB), decision tree (DT), and multi-layer perceptron (MLP)). Performance metrics (accuracy, precision, recall, F1 score, and area under the curve (AUC)) were computed. Results: ViT Base Patch16 384 with MLP achieved the highest accuracy (0.8462 ± 0.0330) and AUC (0.914 ± 0.032). Although CNN models like EfficientNetB0 + MLP performed competitively (0.8334 ± 0.0479 accuracy), ViT models demonstrated more balanced and robust performance. Conclusions: ViT models outperformed CNNs in classifying open apex, suggesting their integration into dental radiologic decision support systems. Future studies should focus on multi-center and multimodal data to improve generalizability.

Detection of optic disc in retinal images is an important step in disease diagnosis and patient follow-up. Optic disc detection is a Preprocess step for the diagnosis of many diseases, such as glaucoma and DP (Diabetic Retinopathy), which are vital for the eye. In this study, a hybrid approach is proposed for fully automatic segmentation of Optic Disc (OD). The first phase of the study consists of OD localization. The OD localization coordinate obtained in the study was used as an input to the method of region segmentation, which is a semi-automatic segmentation method, used as an initial point for OD segmentation. In this way, fully automatic segmentation is done. The success of the hybrid approach was evaluated according to both localization and segmentation. Performance of the study was evaluated with 30 images according to Dice and Jaccard Similarity. As a result of the evaluation have been obtained respectively % 92 and % 87 success.

The proliferation of Internet of Things (IoT) devices in healthcare has led to a continuous flow of delay-sensitive data such as electrocardiography (ECG) and glucose measurements. Traditional cloud-based architectures may not be able to process this data within clinically acceptable response times. Fog computing addresses this limitation by extending computation to regions closer to the data source. This presents a complex multi-objective module deployment problem. In this study, four module deployment strategies (First-In, First-Out (FCFS), Edge-Oriented, Genetic Algorithm (GA), and Particle Swarm Optimization (PSO)) on a four-layer fog-cloud architecture modeled in iFogSim are compared in terms of latency, energy consumption, makespan, and cost values calculated in the simulation environment. Two different topologies are considered in this study. The first is a homogeneous topology where all fog nodes share the same resources and gateways, and the second is a heterogeneous topology where the processing capacities of the end devices differ. Each algorithm was evaluated under four patient load scenarios using a weighted multi-objective fitness function incorporating latency, power consumption, and network usage in a homogeneous topology. Using another fitness function with some penalty parameters in a heterogeneous topology, each algorithm was evaluated under four patient loads. Experimental results show that GA provides lower latency at medium loads, while PSO exhibits more competitive latency performance at high loads. However, the cost difference between the two algorithms remains negligible in all scenarios. FCFS exhibits the worst overall performance, demonstrating the inadequacy of cloud-based strategies alone for real-time healthcare. All algorithms reach saturation above 30 patients, indicating that fog layer hardware capacity, rather than algorithm selection, becomes the dominant performance bottleneck.