دانشگاه علوم پزشکی کرمان Journal of Health and Biomedical Informatics 2423-3870 12 3 2025 12 1 Classification of Schizophrenia Patients using EEG- based Functional Connectivity Map 216 235 FA Mohammad Reza Tamkin M.Sc in Electrical Engineering, Department of Electrical Engineering, Shahid Bahonar University of Kerman, Kerman, Iran Hoda Jalalkamali Assistant Professor, PhD in Cognitive Neuroscience, Computer Engineering Department, Higher Education Complex of Zarand, Shahid Bahoner University of Kerman, Kerman, Iran Hossein Nezam Abadi-pour Professor, Department of Electrical Engineering, Shahid Bahonar University of Kerman, Kerman, Iran Introduction: Functional brain connectivity derived from electroencephalography (EEG) has been widely used to investigate dynamic cognitive processes, information flow in the brain, and the diagnosis of neurological and psychiatric disorders such as schizophrenia. Schizophrenia is a complex disorder characterized by disruptions in functional and attentional brain networks. Accurately examining these alterations requires analytical approaches with high temporal resolution and network-based modeling capabilities. The aim of this study was to analyze patterns of effective brain connectivity in patients with schizophrenia compared with healthy individuals using EEG data and to apply advanced deep learning models for group classification. Method: This study analyzed EEG-based brain connectivity networks of 17 healthy participants and 19 patients with schizophrenia while they performed a task involving the comparison of unilateral and bilateral visual processing. After data processing, functional brain graphs were constructed, and four graph-theoretical metrics including node strength, global efficiency, clustering coefficient, and modularity index were calculated. These features were then used to classify healthy participants and patients with schizophrenia using a deep learning model based on a Long Short-Term Memory (LSTM). Results: The performance of the classification model was evaluated using accuracy, precision, recall, and F-score metrics. The proposed method achieved an accuracy of 94%, outperforming previous studies that classified patients with schizophrenia and healthy participants based on functional connectivity features. Conclusion: The results demonstrate the effectiveness of an LSTM-based recurrent deep learning model in distinguishing patients with schizophrenia from healthy individuals using EEG data. Additionally, the significance of features obtained from the functional brain graph and the task utilized is emphasized. The high classification accuracy suggests that the observed deficits in attention-related brain networks in patients with schizophrenia particularly during unilateral and bilateral visual processing. These deficits represent reliable neural markers differentiating patients with schizophrenia from healthy individuals.

دانشگاه علوم پزشکی کرمان Journal of Health and Biomedical Informatics 2423-3870 12 3 2025 12 1 Predicting Cardiovascular Diseases from EEG Signals Using YOLO Network Composition and XGBoost Algorithm 236 249 FA Alireza Asghari Kiadeh M.Sc. Student, Department of Computer Engineering, FSh.C., Islamic Azad University, Fouman, Iran Sara Motamed Associate Professor, Department of Computer Engineering, FSh.C., Islamic Azad University, Fouman, Iran Introduction: Cardiovascular diseases continue to be one of the leading causes of mortality worldwide, making early detection crucial for reducing complications and fatalities. Traditional diagnostic methods often depend on the manual interpretation of cardiac signals, which can be time-consuming and reliant on the expertise of clinicians. However, advances in deep learning and data mining have facilitated the automated analysis of medical signals and the detection of hidden patterns. This study presents a novel approach to identifying cardiac diseases based on ECG signals. Method: In this study, we developed a hybrid deep learning model that utilizes the fast version of the YOLO network to extract mid-level features from ECG signals, combined with the XGBoost algorithm to improve classification performance. ECG signals from the MIT-BIH Arrhythmia Database were first extracted and preprocessed. These signals were then input into the YOLO network to generate feature vectors. The outputs of the network were subsequently fed into the XGBoost algorithm for final classification using an ensemble of multiple weak decision trees. Finally, we compared the proposed model with established conventional methods. Results: The results demonstrated that the YOLO network effectively extracts key features from ECG signals, and its integration with XGBoost significantly enhances overall model accuracy. The proposed model outperformed baseline methods, including simple neural networks and support vector machines (SVMs). Evaluation on the MIT-BIH Arrhythmia Database revealed substantial improvements in classification accuracy, as well as enhanced sensitivity and specificity. These findings indicate that combining deep learning with boosting algorithms provides an efficient approach for medical signal analysis. Conclusion: The YOLO-XGBoost hybrid model offers an accurate and innovative approach for detecting cardiac diseases from ECG signals. In addition to enhancing classification accuracy, this method is well-suited for implementation in clinical decision support systems and can serve as an efficient tool for patient screening. Future work could involve validating the model with larger and more diverse datasets and integrating it into intelligent medical systems.

دانشگاه علوم پزشکی کرمان Journal of Health and Biomedical Informatics 2423-3870 12 3 2025 12 1 Development of an Ontology Model for the Blood and Blood Products Supply and Demand Chain in Iran 250 264 FA Afshin Sarafinejad Associate Professor, PhD in Medical Informatics, Medical Informatics Research Center, Institute for Futures Studies in Health, Kerman University of Medical Sciences, Kerman, Iran. Mahdie ShojaeiBaghini Assistant Professor, PhD in Health Information Management, Medical Informatics Research Center, Institute for Futures Studies in Health, Kerman University of Medical Sciences, Kerman, Iran Alihasan Ahmadipour PhD Candidate in Medical Informatics, Faculty of Management and Medical Information Sciences, Kerman University of Medical Sciences, Kerman, Iran Introduction: Ontology, as a formal method for accurately representing concepts, entities, and their relationships, plays a vital role in modeling complex processes. Despite numerous studies in this field, the application of ontology-based models in Iran’s blood transfusion chain remains limited. This study aims to develop an ontology model for blood supply and demand in Iran to support decision-making, facilitate information exchange, and enhance efficiency and safety in the blood transfusion chain by integrating relevant concepts and processes. Method: The METHONTOLOGY methodology was used to develop the ontology model, which includes the stages of specification, knowledge acquisition, conceptualization, integration, implementation, evaluation, and documentation. The ontology was implemented in Protégé (version 5.6.3), and its quality was assessed using the FOCA methodology. Additionally, two diagrams, a data flow diagram and a swimlane diagram, were created to enhance the conceptualization of processes. Results: Two ontologies were developed: one for hospitals and another for the Blood Transfusion Organization. The hospital ontology comprised 26 classes and subclasses, 46 object properties, 43 data properties, and 644 axioms. In contrast, the Blood Transfusion Organization ontology included 11 classes and subclasses, 24 object properties, 25 data properties, and 318 axioms. After defining the components and rules, graphical models of both ontologies were created. The quality evaluation results indicated high quality, with scores of 0.995 for the hospital model and 0.951 for the Blood Transfusion Organization model. Conclusion: Designing and developing an ontology model for the blood supply and demand chain enhances the understanding of domain processes, facilitates data sharing, and supports knowledge documentation and transfer to users. Effective ontology development requires a comprehensive understanding of the activities, tasks, concepts, and relationships within the target domain. Furthermore, ontology evaluation is crucial for identifying conceptual deficiencies, optimizing knowledge structures, and improving organizational processes.

دانشگاه علوم پزشکی کرمان Journal of Health and Biomedical Informatics 2423-3870 12 3 2025 12 1 Exploring the Experiences of Elderly Burn Victims with Kinect-Based Motor Rehabilitation: A Qualitative Study 265 276 FA Kolsoum Deldar Assistant Professor, PhD in Medical Informatics, Department of Health Information Technology, School of Allied Medical Sciences, Shahrud University of Medical Sciences, Shahroud, Iran Seyed Hasan Tavousi Associate Professor, MD, Department of General Surgery, School of Medicine, Mashhad University of Medical Sciences, Mashhad, Iran Farnoosh Khodabandeloo MD, Department of Surgery, School of Medicine, North Khorasan University of Medical Sciences, Bojnourd, Iran Mohsen Rakhsha MD, Department of Surgery, School of Medicine, North Khorasan University of Medical Sciences, Bojnourd, Iran Razieh Froutan Associate Professor, PhD in Nursing, Nursing and Midwifery Care Research Center, Mashhad University of Medical Sciences, Mashhad, Iran Introduction: Elderly burn victims face physical limitations that hinder rehabilitation. Technologies like Kinect motion systems have been used to boost patient participation; however little is known about their experiences and perceptions. This study explores elderly burn patients’ experiences with Kinect-based motor rehabilitation. Method: This qualitative study was conducted in the spring of 2025 using a conventional content analysis approach. Participants were purposively selected from hospitalized older adults with burn injuries admitted to the Burn Unit of Imam Reza Hospital in Mashhad, Iran, all of whom had experience participating in a Kinect‑based technology‑assisted rehabilitation program. Data were collected through semi‑structured interviews. Following transcription, the data were managed in Microsoft Excel (version 2019) and systematically coded and categorized. Results: Analysis of 18 individual interviews and one focus group discussion yielded 215 initial codes, organized into four main categories and 12 subcategories. The four themes were: "from confusion to adaptation", "enhanced motivation and engagement in the treatment process", "perceived independence and self- efficacy", and "sense of safety and comfort with technology". Findings revealed that while the use of technology in rehabilitation was initially associated with feelings of unfamiliarity and anxiety, adaptation improved over time through increased selfconfidence and support from healthcare professionals and family members. Technology enhanced patients’ motivation, active participation, and adherence to rehabilitation exercises, making the process more enjoyable. Additionally, increased independence, improved selfefficacy, and reduced reliance on others emerged as key outcomes. Ultimately, the technology fostered a sense of safety and control, reduced stress and pain during exercises, and improved the overall rehabilitation experience. Conclusion: Kinect-based rehabilitation can enhance elderly burn victims’ participation, independence, and satisfaction; however, individual and contextual challenges must be addressed for effective use in this population.

دانشگاه علوم پزشکی کرمان Journal of Health and Biomedical Informatics 2423-3870 12 3 2025 12 1 A Narrative Critical Review of AI Innovations in Toxicity Prediction and Personalized Healthcare 277 293 FA Sakineh Khanamani Falahatipour Assistant Professor, Department of Comparative Biosciences, Faculty of Veterinary Medicine, University of Tehran, Tehran, Iran. Motahareh Soltani Researcher, Pistachio Safety Research Center, Rafsanjan University of Medical Sciences, Rafsanjan, Iran Somayyeh Karami-Mohajeri Associate Professor, Pharmaceutical Research Center, Institute of Neuropharmacology, Faculty of Pharmacy, Kerman University of Medical Sciences, Kerman, Iran Amirreza Ghasemloo Faculty of Mathematical Sciences and Computer, Kharazmi University, Tehran, Iran Soudeh Khanamani Falahati-pour Associate Professor, Pistachio Safety Research Center, Rafsanjan University of Medical Sciences, Rafsanjan, Iran Salimeh Khanamani Falahatipour Researcher, Computer Engineering Department, Faculty of Engineering, Shahid Bahonar University, Kerman, Iran Zohreh Oghabian Assistant Professor, Pharmaceutical Research Center, Institute of Neuropharmacology, Faculty of Pharmacy, Kerman University of Medical Sciences, Kerman, Iran Introduction: Predictive toxicology utilizing artificial intelligence (AI) enables rapid, accurate, and cost-effective assessment of the toxicity of chemicals, pharmaceuticals, and environmental pollutants. By reducing reliance on traditional in vitro and animal testing, this field plays a critical role in drug development, personalized medicine, and environmental health. This study reviews recent advances in AI applications in predictive toxicology and examines the associated opportunities and challenges. Method: This narrative critical review was conducted through a systematic search of scientific databases, including PubMed, Scopus, ScienceDirect, and Google Scholar. Keywords such as "Artificial Intelligence", "Predictive Toxicology," and "Personalized Medicine" were used to identify relevant articles. Articles were selected based on inclusion criteria (focusing on machine learning methods, published from 2000 onward, and full-text availability) and exclusion criteria (irrelevant articles or those lacking experimental data), and were critically evaluated. Results: Machine learning and deep learning algorithms have enhanced predictive models such as QSAR, toxicogenomics, and molecular modeling. These technologies have improved the accuracy of toxicity predictions, accelerated compound screening, and reduced the need for animal testing. In personalized medicine, AI predicts drug toxicity and optimizes therapeutic dosages by analyzing genetic profiles. In environmental health, wearable sensor data are analyzed to monitor the pollution effects. However, challenges include limited high-quality data, restricted model interpretability, and regulatory barriers. Conclusion: Artificial intelligence has significant potential to transform predictive toxicology and enhance human and environmental safety. To fully harness this technology, data standardization, the development of explainable AI models, and the establishment of effective regulatory frameworks are essential. Future directions include real-time toxicity monitoring and the integration of AI with emerging technologies such as CRISPR. These approaches can lead to more informed decision-making, reduced uncertainty in risk assessment, and an accelerated transition toward preventive and personalized toxicology.

دانشگاه علوم پزشکی کرمان Journal of Health and Biomedical Informatics 2423-3870 12 3 2025 12 1 Telemedicine-Based Clinical Rounds: A Strategy for Enhancing Patient Safety in Teaching Hospitals 294 297 FA Amin Beigzadeh Ph.D. in Medical Education, Assistant Professor, Education Development Center, Sirjan School of Medical Sciences, Sirjan, Iran. Mahdieh Montazeri Ph.D. in Medical Informatics, Assistant Professor, Department of Health Information Sciences, Faculty of Management and Medical Information Sciences, Kerman University of Medical Sciences, Kerman, Iran The article is a letter to the editor, so it does not have an abstract.