Ugur Sevik

Dr. Ugur Sevik

Department of Computer Science, Karadeniz Technical University, Trabzon, Turkey (Türkiye)
orcid

Research Focus & Metrics

Academic Career Length

19Years (2008-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 April 2014 in Journal of Biomedical Optics
Verified
Journal Article
Closed access
PDF file
Ugur SevikUgur Sevik
Ugur Sevik
CKCemal Köse
Cemal Köse
TBTolga Berber
Tolga Berber
HEHidayet Erdöl
Hidayet Erdöl
Retinal image quality assessment (IQA) is a crucial process for automated retinal image analysis systems to obtain an accurate and successful diagnosis of retinal diseases. Consequently, the first step in a good retinal image analysis system is measuring the quality of the input image. We present an approach for finding medically suitable retinal images for retinal diagnosis. We used a three-class grading system that consists of good, bad, and outlier classes. We created a retinal image quality dataset with a total of 216 consecutive images called the Diabetic Retinopathy Image Database. We identified the suitable images within the good images for automatic retinal image analysis systems using a novel method. Subsequently, we evaluated our retinal image suitability approach using the Digital Retinal Images for Vessel Extraction and Standard Diabetic Retinopathy Database Calibration level 1 public datasets. The results were measured through the F1 metric, which is a harmonic mean of precision and recall metrics. The highest F1 scores of the IQA tests were 99.60%, 96.50%, and 85.00% for good, bad, and outlier classes, respectively. Additionally, the accuracy of our suitable image detection approach was 98.08%. Our approach can be integrated into any automatic retinal analysis system with sufficient performance scores.

23 September 2008 in Journal of Medical Systems
Verified
Journal Article
Closed access
PDF file
CKCemal Köse
Cemal Köse
Ugur SevikUgur Sevik
Ugur Sevik
OGOkyay Gençalioğlu
Okyay Gençalioğlu
CICevat İkIbaş
Cevat İkIbaş
TKTemel Kayıkıçıoğlu
Temel Kayıkıçıoğlu

14 August 2019 in IET Image Processing
Verified
Journal Article
Open access
PDF file
Ugur SevikUgur Sevik
Ugur Sevik
EKErdinç Karakullukçu
Erdinç Karakullukçu
TBTolga Berber
Tolga Berber
YAYeşim Akbaş
Yeşim Akbaş
STSerdar Türkyılmaz
Serdar Türkyılmaz
Current standard of burn wound evaluation is based on digital photography of wounds examined by a burn specialist. Due to subjectivity of this approach, automated burn wound analysis systems are being developed by researchers. Those systems should contain three major components: segmentation of burn images, feature extraction, and classification of segmented regions into healthy skin, burned skin, and background. The first purpose of this study is to examine various methods in each of these steps and to achieve the best combination. Comparing the performance of segmentation‐based classification approach versus deep learning is the second goal of the study. SegNet‐based semantic segmentation was implemented as a deep learning approach. The best combination to successfully classify the images into skin, burn, and background regions was found to be the fuzzy c‐means algorithm for the segmentation part, and a multilayer feed‐forward artificial neural network trained by the back‐propagation algorithm for the classification part. Having an F ‐score of 74.28% in the classification of images captured without a protocol, the proposed scheme managed to achieve similar results with deep learning, which had an F ‐score of 80.50%.

5 August 2025 in Diagnostics
Verified
Journal Article
Open access
PDF file
Ugur SevikUgur Sevik
Ugur Sevik
Onur MutluOnur Mutlu
Onur Mutlu
Background/Objectives: The diagnosis of pediatric dental conditions from panoramic radiographs is uniquely challenging due to the dynamic nature of the mixed dentition phase, which can lead to subjective and inconsistent interpretations. This study aims to develop and rigorously validate an advanced deep learning model to enhance diagnostic accuracy and efficiency in pediatric dentistry, providing an objective tool to support clinical decision-making. Methods: An initial comparative study of four state-of-the-art YOLO variants (YOLOv8, v9, v10, and v11) was conducted to identify the optimal architecture for detecting four common findings: Dental Caries, Deciduous Tooth, Root Canal Treatment, and Pulpotomy. A stringent two-tiered validation strategy was employed: a primary public dataset (n = 644 images) was used for training and model selection, while a completely independent external dataset (n = 150 images) was used for final testing. All annotations were validated by a dual-expert team comprising a board-certified pediatric dentist and an experienced oral and maxillofacial radiologist. Results: Based on its leading performance on the internal validation set, YOLOv11x was selected as the optimal model, achieving a mean Average Precision (mAP50) of 0.91. When evaluated on the independent external test set, the model demonstrated robust generalization, achieving an overall F1-Score of 0.81 and a mAP50 of 0.82. It yielded clinically valuable recall rates for therapeutic interventions (Root Canal Treatment: 88%; Pulpotomy: 86%) and other conditions (Deciduous Tooth: 84%; Dental Caries: 79%). Conclusions: Validated through a rigorous dual-dataset and dual-expert process, the YOLOv11x model demonstrates its potential as an accurate and reliable tool for automated detection in pediatric panoramic radiographs. This work suggests that such AI-driven systems can serve as valuable assistive tools for clinicians by supporting diagnostic workflows and contributing to the consistent detection of common dental findings in pediatric patients.

29 October 2025 in Diagnostics
Verified
Journal Article
Open access
PDF file
Ugur SevikUgur Sevik
Ugur Sevik
Onur MutluOnur Mutlu
Onur Mutlu
Background/Objectives: The prevailing paradigm in ophthalmic AI involves siloed, single-disease models, which fails to address the complexity of differential diagnosis in clinical practice. This study aimed to develop and validate a unified deep learning framework for the automated multi-class classification of a wide spectrum of retinal pathologies from fundus photographs, moving beyond the single-disease paradigm to create a comprehensive screening tool. Methods: A publicly available dataset was manually curated by an ophthalmologist, resulting in 1841 images across nine classes, including Diabetic Retinopathy, Glaucoma, and Healthy retinas. After extensive data augmentation to mitigate class imbalance, three pre-trained CNN architectures (ResNet-152, EfficientNetV2, and a YOLOv11-based classifier) were comparatively evaluated. The models were trained using transfer learning and their performance was assessed on an independent test set using accuracy, macro-averaged F1-score, and Area Under the Curve (AUC). Results: The YOLOv11-based classifier demonstrated superior performance over the other architectures on the validation set. On the final independent test set, it achieved a robust overall accuracy of 0.861 and a macro-averaged F1-score of 0.861. The model yielded a validation set AUC of 0.961, which was statistically superior to both ResNet-152 (p < 0.001) and EfficientNetV2 (p < 0.01) as confirmed by the DeLong test. Conclusions: A unified deep learning framework, leveraging a YOLOv11 backbone, can accurately classify nine distinct retinal conditions from a single fundus photograph. This holistic approach moves beyond the limitations of single-disease algorithms, offering considerable promise as a comprehensive AI-driven screening tool to augment clinical decision-making and enhance diagnostic efficiency in ophthalmology.

3 July 2025 in Preprints.org
Verified
Preprint
Open access
PDF file
Ugur SevikUgur Sevik
Ugur Sevik
Onur MutluOnur Mutlu
Onur Mutlu
Background/Objectives: The diagnosis of pediatric dental conditions from panoramic radiographs is uniquely challenging due to the dynamic nature of the mixed dentition phase, which can lead to subjective and inconsistent interpretations. This study aims to develop and rigorously validate an advanced deep learning model to enhance diagnostic accuracy and efficiency in pediatric dentistry, providing an objective tool to support clinical decision-making. Methods: YOLO object detection models was developed for the automated identification of four conditions: Dental Caries, Deciduous Tooth, Root Canal Treatment, and Pulpotomy. A stringent two-tiered validation strategy was employed: a primary public dataset (n=644 images) was used for training and the comparative selection of YOLOv11x, while a completely independent external dataset (n=150 images) was used for final testing. All annotations across both datasets were validated by a dual-expert team, comprising a radiologist and a pediatric dentist, ensuring high-quality ground truth. Results: On internal validation, YOLOv11x was selected as the optimal model, achieving a superior mean Average Precision (mAP50) of 0.91. More significantly, when evaluated on the independent external test set, the model demonstrated robust generalization with an overall F1-Score of 0.81 and an mAP50 of 0.82. It yielded strong recall rates for therapeutic interventions (Root Canal Treatment: 88%, Pulpotomy: 86%) and clinically relevant rates for other conditions (Deciduous Tooth: 84%, Dental Caries: 79%). Conclusions: Validated through a rigorous dual-dataset and dual-expert process, the YOLOv11x model proves to be an accurate and reliable tool for automated detection in pediatric panoramic radiographs. This work provides strong evidence that AI-driven systems can augment clinical decision-making, enhance diagnostic precision, and ultimately contribute to improved dental healthcare outcomes for children.