Background/Objectives: In this study, a vision transformer (ViT) based ensemble architecture was developed for the classification of normal, benign, and malignant diseases from breast ultrasound images. The breast ultrasound images (BUSI) dataset was used for the implementation of the proposed method. This dataset includes 133 normal, 437 benign, and 210 malignant ultrasound images. Methods: ROI segmentation and image preprocessing were applied to the dataset to select only the tumor region and use it in the model. Thus, a better performance was achieved using only the lesion regions. Image augmentation was performed using the Albumentations library to increase the number of images. Feature extraction was performed on the obtained images using three ViT-based models (ViT-Base, DeiT, ViT-Small). The purpose of using three different models is to achieve high accuracy. The extracted features were classified using a multilayer perceptron (MLP). Training was performed using 10-fold stratified cross-validation. Results: The purpose of stratified cross-validation is to include a certain number of images from all three classes in each cross-validation proposed model provided 96.2% precision and 86.3% recall for the benign class and 92.9% recall and 76.4% precision for the malignant class. The normal class achieved 100% success. The area under the curve (AUC) values were 0.97, 0.96, and 1.00 for benign and malignant tumors, respectively, and 1.00 for normal tumors. Conclusions: The ROI-based ViT + MLP + Ensemble architecture provided higher accuracy and explainability compared to traditional convolutional neural network (CNN) based methods in medical image classification. It demonstrated a stable success, especially in minority classes, and presented a potential, reliable, and flexible solution in clinical decision support systems.

Metals are one of the most important building materials of modern times. Especially the production and metalworking process of flat metal sheets is very sensitive. Control of the manufacturing process affects not only the intermediate products but also the quality of final products. Early detection of defects on steel plate surfaces is an important task in industrial production. Process control and mistake detection have traditionally been done manually by experts. However, this method is not proper in terms of both time and cost. With the industrial revolution IR 4.0, machine learning (ML) techniques have been developed to solve fault detection problems in products. This study focuses on developing basic machine learning methods for the detection of six different error classes that may occur during production on steel surfaces. Five standard ML models: LD, KNN, DT, SVM, RF, and deep learning (DNN) model: one-dimensional DNN was developed for the classification problem. The UCI steel plate deformation data set was used as the experimental data set. Five performance criteria: Accuracy, Sensitivity, Specificity, Precision, and F1 value were used to determine the success of the methods. The success rates of LD, KNN, DT, SVM, RF and DNN classification methods were 90.136%, 91.7880%, 93.013%, 93.287%, 95.479%, 96.986%, respectively. The results show the significant impact of the machine learning approach on the steel plate fault diagnosis problem.

In recent years, researchers have conducted many studies on the design and control of prosthesis devices that take the place of a missing limb. Functional ability of prosthesis hands that mimic biological hand functions increases depending on the number of independent finger movements possible. From this perspective, in this study, six different finger movements were given to a prosthesis hand via bioelectrical signals, and the functionality of the prosthesis hand was increased. Bioelectrical signals were recorded by surface electromyography for four muscles with the help of surface electrodes. The recorded bioelectrical signals were subjected to a series of preprocessing and feature extraction processes. In order to create meaningful patterns of motion and an effective cognitive interaction network between the human and the prosthetic hand, fuzzy logic classification algorithms were developed. A five-fingered and 15-jointed prosthetic hand was designed via SolidWorks, and a prosthetic prototype was produced by a 3D printer. In addition, prosthetic hand simulator was designed in Matlab/SimMechanics. Pattern control of both the simulator and the prototype hand in real time was achieved. Position control of motors connected to each joint of the prosthetic hand was provided by a PID controller. Thus, an effective cognitive communication network established between the user, and the real-time pattern control of the prosthesis was provided by bioelectrical signals.