Orhan Er

The primary objective of this study is to assess the effect of systematic dataset augmentation on the accuracy of real-time, vision-based Personal Protective Equipment (PPE) detection systems in occupational environments. The PPEDS-1000 dataset was employed, comprising 1,000 expertly annotated images across four PPE usage categories: worker (W), worker with helmet (WH), worker with vest (WV), and worker with both helmet and vest (WHV). An augmented dataset (PPEDS-2600) was derived via controlled geometric transformations (horizontal and vertical flips), additive Gaussian noise, and Gaussian blur. Each dataset is partitioned using an 80/10/10 train–validation–test split and utilized to train five YOLOv8 model variants (nano through extra-large). The evaluation metrics include precision, recall, F1-score, mean average precision at an IoU threshold of 0.5 (mAP50), and mean average precision averaged over IoU thresholds from 0.5 to 0.95 (mAP50-95). The experimental results demonstrate that augmentation elevates mAP50 from 77.7% on PPEDS-1000 to 94.8% on PPEDS-2600, thereby substantiating the hypothesis that targeted augmentation markedly enhances detection performance. The findings indicate that the present work establishes a rigorous benchmark for real-time PPE compliance monitoring.

Classification of different types of strawberries is an essential problem in agriculture and quality control; however, it presents a challenge to traditional deep learning solutions because of the need for large-scale labeled datasets. In this research, a few-shot learning (FSL) model based on Prototypical Networks was applied to classify four different commercial types of strawberries (A, B, C, and Jumbo), utilizing the Mendeley Strawberry Grading Dataset that includes only 103 labeled images for training. A unique pipeline for training an FSL model was created through a combination of warm-up pre-training, differential learning rate strategy, and 4-way 5-shot episodic training; two backbones (ResNet50 and EfficientNet-B0) were compared under a 5-fold stratified cross-validation. While ResNet50 obtained test accuracy, macro F1-score, and Cohen's Kappa equal to 99.00%, 99.11%, and 98.62% respectively, EfficientNet-B0 achieved 98.00%, 98.33%, and 97.30%. Both models produced an AUC value of 99.80%. Interpretability analysis utilizing GradCAM++ showed that both networks focused on relevant areas of the fruits. This research proved that metric-based few-shot learning can be used successfully as an alternative solution to traditional deep learning methods in case of scarce datasets of agricultural images.

Consistent use of personal protective equipment (PPE) reduces exposure to industrial hazards, yet continuous compliance monitoring remains difficult in fast-moving workplaces. This study evaluates lightweight, real-time PPE detectors using two purpose-built datasets: PPEDS-1000 (1,000 manually annotated real-world images) and PPEDS-2600 (an augmented extension designed to reflect deployment variability). YOLOv13-nano and YOLOv13-small, implemented from the publicly available iMoonLab repository, were trained as untuned baselines, while optimization focused on YOLOv8 due to its mature tooling and scalable lightweight variants. Optuna-based hyperparameter optimization was restricted to YOLOv8-nano and YOLOv8-small to preserve deployment feasibility, tuning key training factors such as learning rate, batch size, and weight decay with validation mAP50 as the objective. Compared with pre-optimization baselines, the single-run best configurations achieved absolute gains of up to 2.5 percentage points in mAP50 and 3.2 percentage points in mAP50–95. However, an exploratory repeated-run analysis suggests that the stability of these gains is dataset-dependent. On the smaller PPEDS-1000 dataset, tuning benefits appeared seed-sensitive, whereas on the larger and augmented PPEDS-2600 dataset, tuning showed more consistent behavior together with lower variance across random initializations. Overall, the findings indicate that structured hyperparameter tuning is promising for deployment-oriented PPE monitoring, but its ability to yield stable gains appears to depend on adequate dataset scale and representative variability.

Background/Objectives: This paper aims to design an explainable and accurate ML framework to support the automatic classification of Carpal Tunnel Syndrome (CTS) severity from structured patient data. Methods: For the experiment, an open-source dataset of 1037 samples was used. Following stratified partitioning, 305 samples were held out as the test set; the remaining training set (n = 732) was augmented to 1216 balanced samples via ADASYN, yielding an 80/20 train/test ratio relative to the final dataset (n = 1521). In order to solve the problem of imbalance associated with CTS cases of moderate and severe severity, the Adaptive Synthetic Sampling (ADASYN) technique was employed. The model’s predictive capacity was increased by means of feature engineering methods, such as polynomial transformations and clinically relevant interactions. Specifically, four ensemble learning models (XGBoost, Random Forest, LightGBM, and CatBoost) were optimized and ensembled with the use of a stacking approach with a base algorithm of LightGBM. The explainability of the model was ensured through SHAP and LIME analysis. Results: As a result, the stacking ensemble was able to reach a test accuracy of 91.15%, an F1-score of 91.13%, and an ROC-AUC of 0.9708. The proposed ensemble performed superiorly compared to any other individual algorithm while having stable performance across all severity categories. Conclusions: Through the explainability analysis, it was observed that such a classification model relies on important clinically relevant predictors, including cross-sectional area (CSA), duration of symptoms, pain level measured by the numeric rating scale of pain (NRS), and palmar bowing (PB).

Aim: Follicular quality is a key determinant of success in assisted reproductive technologies, directly affecting outcomes such as fertilization, embryo development, implantation, and live birth rates. However, conventional assessment of cumulus-oocyte complexes relies on subjective morphological evaluation, introducing variability and reducing consistency in clinical decision-making. Materials and Methods: A comparative evaluation of various pre-trained deep learning architectures—including both convolutional neural networks and transformer-based models—was conducted for the automated morphological grading of bovine cumulus-oocyte complexes into four quality categories (Grade A–D). A dataset of 1,400 annotated images of cumulus-oocyte complexes, enhanced through data augmentation techniques to increase image diversity, was used for model training and validation. Results: Among the tested architectures, Xception41 (a variant of convolutional neural networks) and Swin Transformer (a transformer-based model) achieved the highest performance, with test accuracies of 74.75% and 73.25%, and macro F1-scores of 0.75 and 0.74, respectively. While both models performed well in grading cumulus-oocyte complexes with distinct morphological features (Grades 3 and 4), classification accuracy decreased for the more subtle differences between Grades 1 and 2. Furthermore, most models exhibited signs of overfitting under the current training configuration. Conclusion: This study demonstrates the potential of deep learning-based approaches to standardize and enhance the efficiency of cumulus-oocyte complexes evaluation in assisted reproductive technologies. Further optimization is needed to improve model generalization and to address challenges in grading morphologically similar follicular structures.

This study details the development of a remote patient monitoring system with a primary focus on a novel, customized Deep Neural Network (DNN) for arrhythmia detection. The system integrates hardware for real-time data collection from biomedical sensors, where IoT-based sensor data is collected and encrypted in a central database for subsequent analysis. The novelty of the work lies in the proposed AI-based software component rather than the hardware assembly, which utilizes accessible components. The developed system is designed to function as a decision support system for healthcare personnel, providing necessary information and alerts through mobile and desktop interfaces. Data obtained from the patient is classified using the proposed deep learning method, and a detailed summary is presented. The customized DNN-based model demonstrated a test accuracy of 99.94%, with a recall of 99.92% and a precision of 99.57%, results which indicate a strong potential for clinical application due to very low false positive and false negative rates. Based on this high accuracy, the model’s outputs have been integrated into user-friendly interfaces to assist healthcare personnel. It is therefore suggested that the patient monitoring system, featuring this high-performance classification model, has the potential to contribute to the early and more reliable detection of significant diseases such as heart abnormalities and arrhythmia.

This chapter provides the general framework for a basic introduction to non-clinical applications of artificial intelligence (AI) in the healthcare industry. All activities related to patient care that are not observable in the disease treatment process can be referred to as non-clinical healthcare activities. The healthcare industry appears as a top priority in almost every subject in both advanced and emerging countries around the world. Therefore, this book is planned with the aim of closely examining and contributing to the dynamics of the healthcare sector, which has such a strategic importance and social impact. In this respect, Volume 1 covers non-clinical applications of AI and, as a next step, Volume 2 will provide information on the impact of AI on clinical implications. In this context, a wide range of topics will be described, from health big data and the characteristics of the health big data production process to key processes related to health service delivery, quality and accreditation. While the health industry is large in scale, it also has a very complex structure in terms of its operations, services and technologies. For this reason, it is one of the sectors expected to internalize and exploit technological advances most quickly. AI, which is contributing to revolutionary innovations in technological development, is spreading widely through intelligent tools with different algorithms based on machine learning and deep learning techniques. While the applications of artificial intelligence have benefits, they also pose challenges. The topics discussed in this book attempt to open the discussion on all aspects of AI with a balanced approach based on the literature. It provides value-added recommendations for public health management that should be considered in the development of health policies needed to improve service standards and quality, especially in the healthcare industry and the healthcare supply chain process.

Healthcare and medical science are inherently dependent on technological advances and innovations for improved care. In recent times we have witnessed a new drive in implementing these advances and innovations through the use of Artificial Intelligence, in both clinical and non-clinical areas. The set of 2 volumes aims to make available the latest research and applications to all, and to present the current state of clinical and non-clinical applications in the health sector and areas open to development, as well as to provide recommendations to policymakers. This volume covers non-clinical applications. The chapters covered in this book have been written by professionals who are experts in the healthcare sector and have academic experience.