Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Mustafa Koyun has a total of 30 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Patterns of Computed Tomography Use, Diagnostic Yield and Radiation Dose in Pediatric Trauma Patients: A Single-Center Retrospective Cross-Sectional Study

    by Mustafa Koyun, Humeyra Sonmezoglu, Bunyamin Ece, Miraç Koç

    Background/Objectives: Computed tomography (CT) is central to pediatric trauma imaging but involves radiation exposure. This study evaluated CT use patterns, diagnostic yield, and radiation dose relative to diagnostic reference levels (DRLs). Methods: This single-center retrospective cross-sectional study included patients younger than 18 years who underwent CT for blunt trauma in 2025. Findings were classified as major pathology positive or negative. CTDIvol and DLP were retrieved from the picture archiving and communication system, and institutional medians were compared with registry-based and European DRLs. Multivariable logistic regression and generalized estimating equations were used. Results: Among 1686 patients (66.3% male; median age, 9.92 years; 2209 examinations), major pathology was detected in 9.0% of patients (152/1686) and 7.2% of examinations (160/2209). No major pathology was found on single-region spine CT. Yield was lowest at 1–<5 years (1.9%) and highest at 15–<18 years (16.2%); older age and multiple CT examinations were independently associated with major pathology. Median cumulative DLP per patient was 583.0 mGy·cm. Institutional median DLP exceeded phantom-matched registry DRLs in 5 of 22 adequately sampled region–age groups, mainly abdominal CT at ≥10 years; head CT medians exceeded the DRL only in infants. Combined chest–abdomen CT, mostly referenced to the 16 cm phantom, was excluded. Conclusions: Low diagnostic yield in several regions and doses above reference levels in selected regions identify areas for optimizing pediatric trauma CT. Whether decision rules, selective imaging, and protocol optimization can reduce exposure without compromising diagnostic safety should be assessed prospectively.

  • Open Access

    Atlas-Assisted Bone Age Estimation from Hand–Wrist Radiographs Using Multimodal Large Language Models: A Comparative Study

    by Erdem Ozkan, Mustafa Koyun

    Background/Objectives: Bone age assessment is critical in pediatric endocrinology and forensic medicine. Although recently developed multimodal large language models (LLMs) show potential in medical imaging, their diagnostic performance in bone age determination has not been sufficiently evaluated. This study evaluates the performance of four multimodal LLMs (ChatGPT-5, Gemini 2.5 Pro, Grok-3, and Claude 4 Sonnet) in bone age determination using the Gilsanz–Ratib (GR) atlas. Methods: This retrospective study included 245 pediatric patients (109 male, 136 female) under the age of 18 who underwent left wrist radiography. Each model estimated bone age using the patient’s radiograph and GR atlas as reference (atlas-assisted prompting). Bone age assessments made by an experienced radiologist using the GR atlas were evaluated as the reference standard. Performance was assessed using mean absolute error (MAE), intraclass correlation coefficient (ICC), and Bland–Altman analysis. Results: ChatGPT-5 demonstrated statistically superior performance, with an MAE of 1.46 years and ICC of 0.849, showing the highest alignment with the reference standard. Gemini 2.5 Pro showed moderate performance, with an MAE of 2.24 years; Grok-3 (MAE: 3.14 years) and Claude 4 Sonnet (MAE: 4.29 years) had error rates that were too high for clinical use. Conclusions: Significant performance differences exist among multimodal LLMs, despite atlas-supported prompting. Only ChatGPT-5 qualified as “clinically useful,” demonstrating potential as an auxiliary tool or educational support under expert supervision. Other models’ reliability remains insufficient.


  • Open Access

    Atlas-Assisted Bone Age Estimation from Hand–Wrist Radiographs Using Multimodal Large Language Models: A Comparative Study

    by Erdem Ozkan, Mustafa Koyun

    Background/Objectives: Bone age assessment is critical in pediatric endocrinology and forensic medicine. Although recently developed multimodal large language models (LLMs) show potential in medical imaging, their diagnostic performance in bone age determination has not been sufficiently evaluated. This study evaluates the performance of four multimodal LLMs (ChatGPT-5, Gemini 2.5 Pro, Grok-3, and Claude 4 Sonnet) in bone age determination using the Gilsanz-Ratib (GR) atlas. Methods: This retrospective study included 245 pediatric patients (109 male, 136 female) under age 18 who underwent left wrist radiography. Each model estimated bone age using the patient's radiograph and GR atlas as reference (atlas-assisted prompting). Bone age assessments made by an experienced radiologist using the GR atlas were evaluated as the reference standard. Performance was assessed using mean absolute error (MAE), intraclass correlation coefficient (ICC), and Bland-Altman analysis. Results: ChatGPT-5 demonstrated statistically superior performance with MAE of 1.46 years and ICC of 0.849, showing highest alignment with the reference standard. Gemini 2.5 Pro showed moderate performance with MAE of 2.24 years; Grok-3 (MAE: 3.14 years) and Claude 4 Sonnet (MAE: 4.29 years) had error rates too high for clinical use. Conclusions: Significant performance differences exist among multimodal LLMs despite atlas-supported prompting. Only ChatGPT-5 qualified as "clinically useful," demonstrating potential as an auxiliary tool or educational support under expert supervision. Other models' reliability remains insufficient.

  • Open Access

    The Predictive Role of the Triglyceride/High-Density Lipoprotein Ratio and the Triglyceride–Glucose Index, Along with Anthropometric Measurements, in Diagnosing Non-Alcoholic Fatty Liver Disease in Obese Kids and Juveniles, and the Evaluation of Novel Cardiovascular Risk Markers in Pediatric NAFLD

    by Emrah Çığrı, Funda Çatan İnan, Sedat Gülten, Mehmet Akif Bildirici, Ayşe Ece Gökkaya, Metin Asıleren, Mustafa Koyun, Bahadır Reis, Merve Esen

    Aim: This current research aims to determine the predictive value of the ratio of triglyceride (TG)/high-density lipoprotein cholesterol (HDL-C), index of triglyceride–glucose (TyG), homeostatic model assessment for insulin resistance (HOMA-IR) score, and anthropometric measurements at the onset of non-alcoholic fatty liver disease (NAFLD) in obese kids and juveniles. It also sought to assess how novel cardiovascular risk markers are affected in obese pediatric patients with NAFLD. Materials and Methods: Between November 2024 and May 2025, a total of 199 pediatric patients were prospectively evaluated, including 150 children with obesity and 49 entirely healthy controls. Two categories of obese patients were created based on whether or not they had non-alcoholic fatty liver disease. These groups were compared with each other and with the control group in terms of HOMA-IR score, index of TyG, ratio of TG/HDL-C, anthropometric parameters (percentage of body fat [BFP], index of body mass [BMI], body fat mass [BFM], waist circumference [WC]), and cardiovascular risk markers. The cutoff values, sensitivity, and specificity of the HOMA-IR score, ratio of TG/HDL-C, anthropometric measurements, and index of TyG in predicting NAFLD were assessed using Receiver Operating Characteristic (ROC) analysis. Results: Obese kids and juveniles with NAFLD had significantly higher TG/HDL-C ratios, TyG indices, HOMA-IR scores, anthropometric measurements, and cardiovascular risk markers than those without NAFLD (p < 0.001). The TG/HDL-C ratio (AUC: 0.936; 81.8% sensitivity, 95.9% specificity) and the TyG index (AUC: 0.912; 81.8% sensitivity, 91.8% specificity) showed strong predictive value for NAFLD, while HOMA-IR and WC were found to be relatively weaker predictors. Conclusions: The index of TyG and ratio of TG/HDL-C are highly effective parameters in predicting NAFLD development in obese kids and juveniles. Those with increased WC and BFP should be closely monitored for NAFLD development. Pediatric patients with NAFLD should be carefully followed up for potential cardiovascular diseases.

  • A Hybrid CNN Framework for Kidney Stone Detection Using Transfer Learning and Feature Fusion

    by Coşku Öksüz, Artun Narter, Bünyamin Ece, Mustafa Koyun, İsmail Taşkent, M. Kemal Güllü

    In this study, a deep learning method for kidney stone detection is proposed. The method utilizes transfer learning by extracting features from a pre-trained ImageNet model. However, unlike traditional transfer learning, which directly applies or fine-tunes a pre-trained model, the proposed approach integrates a custom-designed CNN that operates in parallel with the pre-trained network. The feature maps obtained from both networks are fused to enhance the model's representation power. After this integration, task-specific classification layers are added, and the training process is conducted on both the classification layers and the optimized model. This approach improves the overall performance of the model while providing a more efficient training process. As part of this study, a new dataset was created, consisting of 2166 axial slice images from 241 patients and 2018 axial slice images from 46 healthy individuals. Experiments conducted using EfficientNetV2s, MobileNetV4s, SqueezeNet, and ResNet18-based models revealed that the EfficientNetV2s and MobileNetV4s-based models excelled in terms of accuracy, while the SqueezeNet and ResNet18-based models provided stronger results in terms of interpretability.


  • Open Access

    Comparison of prostate biopsy results with pı-rads categories, prostate volume and prostate specific antigen parameters

    by Mustafa Koyun, İsmail Taşkent, Bünyamin Ece

    Aims: The aim of this study was to compare prostate biopsy results with PI-RADS (Prostate Imaging-Reporting and Data System) categories, prostate volume, and prostate-specific antigen (PSA) parameters. Methods: A total of 367 patients who underwent mpMRI between July 2021 and February 2023 were retrospectively analyzed. Age, total PSA (tPSA), free PSA (fPSA), tPSA/fPSA ratio, PSA density (PSAD, tPSA/prostate volume), and PI-RADS (Prostate Imaging-Reporting and Data System) categories were recorded for each patient. 199 patients were excluded from the study due to insufficient data, and the study was conducted with the remaining 168 patients. The mpMRI examinations were performed using a 1.5 T MRI scanner. PI-RADS categories 1, 2, and 3 obtained from mpMRI were classified as benign, while PI-RADS categories 4 and 5 were classified as malignant. Prostate biopsy was performed as a 12-core biopsy under transrectal ultrasonography (TRUS) guidance. Independent samples t-test and Chi-square test were used to compare the data of histopathologically benign and malignant groups. A p-value <0.05 was considered statistically significant. Results: Of the 168 patients included in the study, 105 (62.5%) were benign and 63 (37.5%) were malignant based on histopathological results. Statistically significant differences were found between malignant/benign groups in mean age (69.9/67.0 years, p=0.002), prostate volume (62.5/93.9 mL, p<0.001), fPSA/tPSA ratio (0.13/0.20, p=0.048), and PSAD (0.25/0.14 ng/mL2, p=0.003); while no significant differences were found in tPSA (13.9/11.5 ng/mL, p=0.419) and fPSA (1.63/1.68 ng/mL, p=0.922) values. A statistically significant difference was found between PI-RADS results of malignant/benign cases (p<0.001). Of the cases classified as malignant according to PI-RADS, 44 (26.2%) cases (18 prostatitis, 26 benign prostate tissue) were histopathologically benign, while 21 (12.5%) cases (8 high-grade prostatic intraepithelial neoplasia, 13 malignant cases) classified as benign according to PI-RADS were histopathologically malignant. In the remaining 103 (61.3%) cases, PI-RADS and histopathology results were concordant. Conclusion: According to our study, while tPSA and fPSA alone were insufficient in detecting prostate malignancy, fPSA/tPSA ratio and PSAD were found to be associated with malignancy. Discordance was detected between PI-RADS categories and histopathological results. The discrepancy between mpMRI and histopathological data may be attributed to the systematic 12-core blind biopsy procedure. MRI-guided biopsy might reduce the discordance between PI-RADS categories and histopathological results. In conclusion, the combined use of PSA parameters and mpMRI data will provide more accurate results in predicting prostate cancer.

  • Open Access

    A rare complication of thoracotomy: thoracodorsal artery pseudoaneurysm

    by Mustafa Koyun, Bahadir Reis

  • Open Access

    Diagnostic Value of T2 Mapping in Sacroiliitis Associated with Spondyloarthropathy

    by Mustafa Koyun, Kemal Niyazi Arda

    Background/Objectives: T2 mapping is a quantitative magnetic resonance imaging (MRI) technique that provides information about tissue water content and molecular mobility. This study aimed to evaluate the diagnostic utility of T2 mapping in assessing sacroiliitis associated with spondyloarthropathy (SpA). Methods: A prospective study examined a total of 56 participants, comprising 31 SpA patients (n = 31) and 25 healthy controls (n = 25), who underwent sacroiliac joint MRI between August 2018 and August 2020. T2 mapping images were generated using multi-echo turbo spin echo (TSE) sequence, and quantitative T2 relaxation times were measured from bone and cartilage regions. Statistical analysis employed appropriate parametric and non-parametric tests with significance set at p < 0.05. Results: The mean T2 relaxation time measured from the areas with osteitis of SpA patients (100.23 ± 7.41 ms; 95% CI: 97.51–102.95) was significantly higher than that of the control group in normal bone (69.44 ± 4.37 ms; 95% CI: 67.64–71.24), and this difference was found to be statistically significant (p < 0.001). No significant difference was observed between cartilage T2 relaxation times in SpA patients and controls (p > 0.05). Conclusions: T2 mapping serves as a valuable quantitative imaging biomarker for diagnosing sacroiliitis associated with SpA, particularly by detecting bone marrow edema. The technique shows promise for objective disease assessment, though larger studies are needed to establish standardized reference values for T2 relaxation times in osteitis to enhance diagnostic accuracy and facilitate treatment monitoring.


  • Erkek Genital Sistem Acil Patolojilerinin Ultrasonografik Görüntüleme Bulguları Eşliğinde Kapsamlı Bir Değerlendirmesi

    by Mustafa Koyun

    Erkek genital sisteminin acil patolojileri, hızlı ve doğru tanı gerektiren klinik durumlardır. Ultrasonografi (US) bu acil patolojilerin tanı sürecinde önemli bir role sahiptir. Doppler teknolojisiyle desteklenen, non-invaziv ve gerçek zamanlı bir değerlendirme imkanı sunan bu görüntüleme yöntemi, testis torsiyonu, testis rüptürü, orşit, epididimit ve penil fraktür gibi acil patolojileri değerlendirmede oldukça etkilidir. Her bir patoloji, kendine özel klinik bulgular (örneğin, ani skrotal ağrı, şişlik), laboratuvar bulgular (lökositoz, kültür sonuçları) ve US görüntüleme bulguları (kan akım değişiklikleri, yapısal anomaliler) ile ortaya çıkar.US, cerrahi açıdan acil patolojileri (torsiyon, fraktür gibi) inflamatuar süreçlerden (orşit, epididimit gibi) ayırmada yüksek hassasiyet sunar ve tedavi planında yol göstericidir. Testis torsiyonunda “spermatik korddaki girdap işareti”, rüptürde tunica albuginea devamlılığında bozulma veya apselerde kanlanmayan hipoekoik lezyonlar gibi bulgular, ilgili tanıyı kesinleştirmede oldukça değerlidir. Bu bölüm, erkek genital sisteminin acil patolojilerinin klinik, laboratuvar ve görüntüleme özelliklerini detaylı bir şekilde ele alarak, US’nin ürolojideki vazgeçilmezliğini vurgulamakta ve hasta sonuçlarını iyileştirme potansiyelini ortaya koymaktadır.