Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Mucahit Karaduman has a total of 76 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    A Comparative Benchmark of Modern Deep Learning Backbones for Multi-Class Periodontal Bone Loss Staging

    by Nurullah Duger, Burak Dagtekin, Furkan Talo, Melek Atille Aydin, Mucahit Karaduman, Muhammed Yildirim, Tuba Talo Yildirim, Tolga Fikret Tozum

    Background/Objectives: Early diagnosis of periodontal disease is critical to prevent tissue destruction and tooth loss. This study evaluates six modern deep learning architectures for four-class radiographic periodontal bone loss grading on panoramic radiographs. Methods: This retrospective study analyzed 1053 images. Two periodontologists independently classified each image according to the most severely affected tooth into four bone loss categories: Healthy, Mild (50%). Six convolutional, transformer, and hybrid backbones (ConvNeXt-Tiny, MaxViT-Tiny, EfficientViT-B1, FocalNet-Tiny, CAFormer-S18, and DaViT-Tiny) were trained on a 70/10/20 stratified split using an NVIDIA Tesla T4 GPU, the AdamW optimizer, Cosine Annealing, and early stopping. Performance was assessed using standard classification metrics, Matthews Correlation Coefficient, Cohen’s and Quadratic Weighted Kappa, Expected Calibration Error, Area Under the Curve (AUC), McNemar’s test, and bootstrap 95% confidence intervals. Results: FocalNet-Tiny achieved the highest numerical scores among the tested models, reaching 83.41% accuracy, 84.70% balanced accuracy, 83.50% macro-F1, and 0.9610 OvR AUC. The p-values obtained from the comparisons of the models were all above 0.05, indicating that the classification performances of the evaluated models are quite similar. Bootstrap analysis confirmed stable prediction metrics for the top-scoring model. Conclusions: This study provides a controlled benchmark of modern deep learning backbones for four-class radiographic periodontal bone loss grading. FocalNet-Tiny achieved the highest numerical metrics and stable bootstrap estimates, yet performed comparably to other contemporary architectures such as ConvNeXt-Tiny. Direct radiographic classification shows promise for automated radiographic pre-screening; however, multi-center external validation and integration with clinical periodontal parameters are required before clinical implementation.

  • Open Access

    Association and Discrimination of Urinary System Stone Disease Using Anthropometric Indices, Lifestyle Behaviors, and Artificial Intelligence: A Multicenter Case–Control Study

    by Mehmet Erhan Aydın, Coşkun Kaya, Aykut Aykaç, Kadir Yildirim, Mucahit Karaduman, Fatih Bicaklioglu, Yiğit Kudret Akyol, Bayram Kolcu, Mehmet Ezer, Ferhat Yakup Suceken, et al.

    Background/Objectives: Urinary system stone disease is common and highly recurrent, and body mass index (BMI) poorly reflects the central adiposity linked to stone risk. We compared BMI with five next-generation anthropometric indices (A Body Shape Index [ABSI], Body Roundness Index [BRI], Waist-to-Height Ratio [WtHR], Conicity Index [ConI], Waist-to-Hip Ratio [WHR]) for their association with stone disease, and developed a hybrid autoencoder-based artificial intelligence model integrating anthropometric, clinical, and lifestyle data. Methods: This multicenter case–control study included 1195 participants (562 stone cases, 633 controls) from six centers in Turkey. Thirty-one variables were grouped into three subgroups (basic measurements, derived indices, clinical/lifestyle data), each processed by a separate autoencoder; combined features were classified with XGBoost, Cubic SVM, Fine KNN, and logistic regression. Each index’s independent association was assessed by quartile-based logistic regression, and discriminative power by ROC/AUC. Results: BMI showed no between-group difference (p = 0.919) or discriminative power (AUC = 0.498). ABSI, BRI, WtHR, ConI, and WHR were all significantly higher in the stone group, showed positive trends across quartiles after full adjustment (all p-trend ≤ 0.001), and outperformed BMI (AUC 0.547–0.564); associations were strongest for ConI (aOR = 2.62) and ABSI (aOR = 2.57). The hybrid AI model performed best with XGBoost (77% accuracy), followed by Fine KNN (71%), logistic regression (69%), and Cubic SVM (68%). Conclusions: Unlike BMI, next-generation indices reflecting central adiposity are independently and consistently associated with stone disease, and a multivariable AI model combining anthropometric, clinical, and lifestyle data reached 77% accuracy in separating stone patients from controls on a single test split. These findings are a hypothesis-generating risk-stratification signal that requires external validation, and they do not support clinical screening.

  • Open Access

    A topology-aware visual fusion framework for patient-level emphysema classification from chest CT images

    by Yesim Eroglu, Muhammed Yildirim, Oguzhan Katar, Esra Yildirim, Mucahit Karaduman, Vedat Tumen, Muhammed Talo, Erdal Akin, Ozal Yildirim, U. Rajendra Acharya

  • Open Access

    Validation Study of ASA Physical Status Prediction Model Developed Through Holistic Integration of Lung Images, Electrocardiogram Data and Clinical Data

    by Kadir Yildirim, Serpil Bayindir, Mucahit Karaduman, Muhammed Yildirim

    Background/Objectives: Objective assessment of patients’ physical health status before surgical interventions is crucial in predicting surgical risks and planning patient management. The American Society of Anesthesiologists Physical Status Classification (ASA-PS) is one of the most widely used scales in this context. It plays a vital role in clinical decision-making processes. However, ASA scoring can vary among clinicians due to its reliance on subjective assessments, making it difficult to standardize. Methods: In this study, a deep learning-based model was developed to classify ASA from patient data automatically. Three different data types were used in this study: lung images, Electrocardiogram (ECG) data, and clinical data. For the proposed ASA prediction model, features were extracted from each data type. Features were extracted from ECG and Computed Tomography (CT) images using five different Convolutional Neural Network (CNN) architectures, while statistical features were derived from the clinical data. The features obtained from the CNN architecture with the highest classification performance were combined with the statistical features. These combined features were then processed using seven different classifiers. Results: The developed model was evaluated with performance metrics such as accuracy, precision, F1 score, and Receiver Operating Characteristic (ROC) curve. As a result, a high accuracy value of 97.30% was obtained. Conclusions: This study aims to make preoperative risk management for ASA classification more objective and consistent, and the results obtained support this aim.

  • Transformer and VOTE Based AI Model for Waste Classification: An Application of Recycling in Smart Cities

    by Mücahit Karaduman, Muhammed Yıldırım

    As industrialization and smart city development activities increase, waste collection, classification, and planning have become important. The recycling process of waste is based on recovering its properties as they are in their natural locations, reducing pollution, and helping to create a sustainable environment. This study aims to improve the performance of deep learning models for converting organic waste into recyclable waste. In the study, a transformer-based hybrid model was proposed for waste classification. The proposed model used DinoV2, ConvNeXtV2, and ViT-B16 as the base models. Feature maps derived using these models were concatenated and then classified into different classifiers. In the last step of our model, the features obtained from different classifiers were subjected to the VOTE hard-voting method. The proposed model achieved 96.64% success in solid waste classification. The VOTE process improved the proposed model's performance from 95.58% to 96.64%. When the features obtained from the transformer-based models used in the study were classified by the classifiers, the highest accuracy, 95.37%, was achieved with the DinoV2 model. The values obtained in the proposed model are important for the future of smart and sustainable cities.

  • Feature-Level Fusion–Based CNN–Transformer–KAN Hybrid Approach for Dental Implant Diameter and Length Detection

    by Furkan Talo, Mucahit Karaduman, Nurullah Duger, Gulsah Karaduman, Muhammed Yildirim, Melek Atille Aydin, Gulucag Giray Tekin, Tuba Talo Yildirim

  • Open Access

    An Attention-Residual Hybrid CNN for CT-Based Multiclass Classification of Alcohol-Related Liver Disease: Differential Diagnosis Against HBV-Related Cirrhosis

    by Ertugrul Karabulut, Mucahit Karaduman, Muhammed Yildirim, Sami Akbulut

    Background: Differentiating alcohol-related liver disease (ARLD) from chronic liver injury caused by other etiologies remains a clinically relevant challenge in cross-sectional imaging. In particular, alcoholic hepatitis and alcoholic cirrhosis may overlap morphologically with HBV-related cirrhosis on CT imaging. To develop and assess an attention-residual hybrid Convolutional Neural Network (CNN) for multiclass CT image classification of alcoholic hepatitis, alcoholic cirrhosis, HBV-related cirrhosis, and living liver donors. Methods: A four-class liver image dataset comprising 5760 CT images from 144 individuals (36 per group) was constructed using images from alcoholic hepatitis, alcoholic cirrhosis, HBV-related cirrhosis, and living liver donors. The dataset was partitioned into training, validation, and test sets at the patient level. Five pretrained CNN architectures, including DenseNet121, ResNet50, MobileNetV3-Large, EfficientNetB0, and ConvNeXt-Tiny, were first fine-tuned and comparatively evaluated. Based on F1-score ranking, DenseNet121 and ConvNeXt-Tiny were selected as the two backbone networks for the proposed hybrid model. The final architecture integrated Attention Pooling, Feature-wise Linear Modulation (FiLM), Multi-head Attention, Gated Linear Units, Residual Connections, and layer normalization to improve feature fusion and contextual representation. Results: The proposed model demonstrated the best overall performance among all evaluated architectures on the test set. It achieved an accuracy of 99.55%, a weighted F1-score of 99.55%, an MCC of 0.9941, and a Cohen’s kappa coefficient of 0.9940. The model also achieved ROC-AUC and PR-AUC values of 100.00% and 99.99%, respectively, together with an NPV of 99.85%. The proposed model’s performance was also balanced across classes. For Alcoholic Cirrhosis, precision, recall, and F1-score were all 99.64%. For Alcoholic Hepatitis, the corresponding values were 100.00%, 98.93%, and 99.46%, respectively, while HBV-related Cirrhosis achieved 99.29% precision, 99.64% recall, and 99.47% F1-score. Living Liver Donors achieved 99.29% precision, 100.00% recall, and 99.64% F1-score. Conclusions: The findings of this exploratory study suggest that routine CT images may contain image-based differences potentially relevant to etiology-oriented classification of diffuse liver disease. Beyond distinguishing ARLD from HBV-related cirrhosis and images from living liver donors, the model also captured image-based differences between major ARLD subgroups, including alcoholic hepatitis and alcoholic cirrhosis. These findings support further investigation of CT-derived image-based differences in larger independent and multicenter datasets.


  • Open Access

    Automatic Detection of Sinus Membrane Thickness Using a Hybrid Deep Learning Model Integrating Multiple Attention Mechanisms

    by Furkan Talo, Nurullah Duger, Burak Dagtekin, Mucahit Karaduman, Muhammed Yildirim, Tuba Talo Yildirim

    Background/Objectives: In dental implant surgeries, automatic detection and classification of maxillary sinus membrane thickening from Cone-Beam Computed Tomography (CBCT) images used for diagnosis and treatment planning, using an artificial intelligence-based system, will both reduce specialists’ error rates and enable faster processing. Therefore, in this study, a unique model was developed to automatically detect maxillary sinus membrane thickening from CBCT images. Methods: The first step of the developed model is data preprocessing. Then, feature extraction is performed using the backbone Convolutional Neural Network (CNN). The obtained features are processed in parallel with Squeeze-and-Excitation (SE), Convolutional Block Attention Module (CBAM), and Efficient Channel Attention (ECA) attention mechanisms. In this stage, channel, spatial, and local relationships are better represented. The combined features are passed to a Multilayer Perceptron (MLP) with dropout to mitigate overlearning, and the final class estimation is performed by a softmax layer. Results: Thanks to the attention mechanisms employed in the developed hybrid model, both the model’s representational power and its classification performance have improved. The model developed to automatically detect and classify maxillary sinus membrane thickening from CBCT images achieved a test accuracy rate of 99.43%. Conclusions: The developed model was compared with pre-trained models accepted in the literature and demonstrated superior performance. The current results demonstrate that the proposed model exhibits high performance in classifying maxillary sinus membrane thickness in a single-center CBCT dataset. However, independent external validation and prospective clinical evaluation are necessary to determine its potential for clinical use.

  • Open Access

    An Attention-Enhanced RegNetY Framework for Detection and Classification of Vertical Misfit in Implant-Supported Restorations: A Retrospective Study

    by Tuba Talo Yildirim, Aybike Cengiz Dagtekin, Nurullah Düger, Ayşe Rençber Kizilkaya, Furkan Talo, Emre Arslan, Mucahit Karaduman, Muhammed Yildirim

    Background/Objectives: The aim of this study is to test different convolutional neural network (CNN) and Transformer-based models to detect and classify vertical misfit at the abutment-prosthesis interface on panoramic radiographs, and to develop a hybrid deep learning model enhanced with attention mechanisms. Methods: A dataset consisting of a total of 566 images, manually classified as 249 ‘fit’ and 317 ‘misfit’ cases by two experts, was created. Images were resized to 224 × 224 and divided into training, validation, and test groups. The deep learning model yielding the most successful results was determined as the backbone; a hybrid model was developed by integrating three different attention modules (SE, CBAM, and ECA) into this structure. Model performance was evaluated using accuracy, precision, sensitivity, and F1 score metrics. Results: CNN-based models (RegNetY-800MF, ConvNeXt-Tiny, EfficientNetV2-S, ResNet50) performed better than Transformer-based models (DeiT, Swin-Tiny) in all metrics. The proposed hybrid model exhibited the highest success among all tested models with a 99.12% accuracy rate. This model reached a 100% precision value in the misfit group and yielded no false positive results. The F1 scores of the hybrid model were recorded as 99.01% for the fit group and 99.21% for the misfit group. Conclusions: The findings of this study demonstrate that attention-enhancing deep learning frameworks have the potential to significantly improve the diagnostic utility of routine panoramic radiographs. It shows that panoramic imaging, when supported by advanced artificial intelligence, can provide valuable diagnostic support in detecting vertical misfit. The developed model has the potential to become a reliable clinical decision support system.


  • Automatic Detection of Dental Implant Brands Using a Hybrid Vision Transformer-Based Deep Learning Model

    by Ayşe Rençber Kızılkaya, Mucahit Karaduman, Harun Kurt, Tuba Talo Yıldırım, Nurullah Düger, Furkan Talo, Muhammed Yıldırım

    Dental implants are commonly used to restore the function and esthetics of missing teeth, and their long-term success depends on appropriate maintenance and follow-up interventions. One of the major clinical challenges is clinicians’ inability to identify the brand of a patient’s previously placed implants, which complicates the selection of compatible prosthetic components and instruments for subsequent treatments. This study aimed to develop a deep learning–based approach to identify dental implant brands automatically. A hybrid model based on Vision Transformer (ViT) architectures was proposed. To enhance efficiency, feature maps from three state-of-the-art ViT models were integrated and then classified with multiple classifiers. The proposed model achieved an accuracy of 87.92%, outperforming widely used benchmark methods in the literature. These findings demonstrate the proposed model's potential to help clinicians accurately identify implant brands, thereby facilitating appropriate prosthetic procedures and improving clinical workflow.