Gülay Güngör

Purpose: To investigate tracheobronchial diameters and tracheobronchial bifurcation angles on computed tomography (CT) in patients with pneumomediastinum.Materials and methods: CT scans of 63 patients with pneumomediastinum (PP) and 63 patients with normal CT findings (PN) were retrospectively evaluated in the Radiology Archives of our hospital. In PP, the extent of mediastinal air on CT images was visually classified into three grades. Thoracic anteroposterior distance (TAPD) and transverse thoracic distance (TTD), anteroposterior (AP) diameters of the trachea and bronchi, tracheal length (TL), right main bronchus angle (RMBA), left main bronchus angle (LMBA), subcarinal angle (SCA), and interbronchial angle (IBA) were measured on CT images in all patients. All measurements were compared between patients with and without pneumomediastinum.Results: A significant sex-related difference was observed in thoracic AP diameter and left main bronchus diameter in the PN group. Among PP, 22.2% (n=14) were grade 1, 30.2% (n=19) were grade 2, and 47.6% (n=30) were grade 3.Tracheal length and left main bronchus diameter were higher in PN patients compared to PP patients. The mean SCA was 75.06±23.40° in the PP group and was significantly greater than in the PN group (p=0.001). LMBA was significantly higher in the PN group compared with the PP group (p=0.003).Conclusion: Individuals with a narrower LMBA, shorter tracheal length, larger thoracic AP diameter, and wider SCA may be at greater risk of developing pneumomediastinum in the setting of barotrauma.

Background/Objective: In recent years, artificial intelligence (AI) has gained increasing prominence in the fields of diagnostic decision-making in medicine. The aim of this study was to compare multidisciplinary team (MDT: rheumatology, pulmonology, thoracic radiology) decisions with single-session plans generated by ChatGPT-4o. Methods: In this cross-sectional concordance study, adults (≥18 years) with confirmed systemic autoimmune rheumatic disease (SARD) and having MDT decisions within the last 6 months were included. The study documented diagnostic, treatment, and monitoring decisions in cases of SARDs by recording answers to six essential questions: (1) What is the most likely clinical diagnosis? (2) What is the most likely radiological diagnosis? (3) Is there a need for anti-inflammatory treatment? (4) Is there a need for antifibrotic treatment? (5) Is drug-free follow-up appropriate? and (6) Are additional investigations required? Consequently, all evaluations were performed with ChatGPT-4o in a single-session format using a standardized single-prompt template, with the system blinded to MDT decisions. All data analyses in this study were conducted using the R programming language (version 4.3.2). An agreement between AI-generated and MDT decisions was assessed using Cohen’s Kappa (κ) statistic where κ (kappa) values represent the level of agreement: 0.80 = almost perfect agreement. These analyses were performed using the irr and psych packages in R. Statistical significance of the models was evaluated through p-values, while overall model fit was assessed using the Likelihood Ratio Test. Results: A total of 47 patients were involved in this study, with a predominance of female patients (61.70%, n = 29). The mean age was 61.74 ± 10.40 years. The most frequently observed diagnosis was rheumatoid arthritis (RA), accounting for 31.91% of cases (n = 15). This was followed by cases of anti-neutrophil cytoplasmic antibody (ANCA)-associated vasculitis, interstitial pneumonia with autoimmune features (IPAF), and sarcoidosis. The analyses indicate a statistically significant level of agreement across all decision types. For clinical diagnosis decisions, agreement was moderate (κ = 0.52), suggesting that the AI system can reach partially consistent conclusions in diagnostic processes. The need for an immunosuppressive treatment and follow-up without medication decisions demonstrated a higher level of concordance, reaching the moderate-to-high range (κ = 0.64 and κ = 0.67, respectively). For antifibrotic treatment decisions, agreement was moderate (κ = 0.49), while radiological diagnosis decisions also fell within the moderate range (κ = 0.55). The lowest agreement—though still moderate—was observed in further investigation required decisions (κ = 0.45). Conclusions: In patients with SARDs with pulmonary involvement, particularly in complex cases, concordance was observed between MDT decisions and AI-generated recommendations regarding prioritization of clinical and radiologic diagnoses, treatment selection, suitability for drug-free follow-up, and the need for further diagnostic investigations.

Aim: This study presents a bibliometric analysis of artificial intelligence (AI)-)-assisted publications in abdominal computed tomography (CT) over the past decade. By examining publication trends, citation patterns, and research collaborations, this study offers insights into the evolving impact of AI in abdominal imaging. Materials and Methods: Data were retrieved from the Web of Science Core Collection using specific search criteria for 2014–2024. Bibliometric analysis was conducted using VOSviewer to generate co-occurrence networks, citation maps, and collaboration patterns. The study included keyword analysis, co-authorship analysis, co-citation analysis, and bibliographic coupling. Results: A significant increase in AI-related publications in abdominal CT has been observed in recent years, with deep learning emerging as the dominant methodology. Citation network analysis identified key studies focused on image reconstruction, segmentation, and radiomics. Collaboration networks highlighted strong international and inter-institutional partnerships, particularly among institutions in the United States, China, and South Korea. Additionally, industry-academic collaborations, notably with GE Healthcare, have contributed to the advancement of AI in abdominal imaging. Conclusions: AI-assisted abdominal CT imaging continues to expand as a critical area of research, demonstrating increasing interdisciplinary collaborations. Deep learning and radiomics have become focal points, influencing clinical decision support and quantitative imaging analysis. Future research should prioritize AI integration into routine radiology practice and explore its clinical effectiveness through large-scale validation studies.

Objective: This study aims to conduct a bibliometric analysis of publications on MRI-based staging of rectal cancer and the applications of artificial intelligence (AI). The study examines key trends in the literature, leading authors, and collaboration networks. Material and Methods: Data were retrieved from the Web of Science (WoS) Core Collection for the period between 2014 and 2024. Citation analysis, co-authorship mapping, bibliographic coupling, and keyword co-occurrence analysis were performed using VOSviewer software. A minimum threshold of two publications and five citations was applied in the analyses. Network, overlay, and density maps were generated. Results: A total of 104 publications were analyzed. The findings indicate that Cui, Cusumano, and Li are among the most frequently cited authors in the field of MRI and AI applications for rectal cancer. Prominent keywords include “rectal cancer,” “magnetic resonance imaging,” “deep learning,” and “machine learning,” reflecting high research activity in these areas. In contrast, terms such as “transfer learning” and “tumor budding” appeared less frequently. Strong collaboration networks were identified, particularly between China and the Netherlands, with GE Healthcare emerging as a key institution in the field. However, the integration of traditional MRI protocols with AI-driven analyses remains limited, highlighting a critical research gap that warrants further exploration. Conclusion: This bibliometric analysis identifies influential themes and leading researchers in MRI and AI applications for rectal cancer. Future research should focus on integrating traditional MRI protocols with AI algorithms and strengthening international collaboration to advance the field.

Aim: This study aims to evaluate the relationship between testicular tissue elasticity and ultrasound (US) histogram analysis obtained from testicular echo signals. The goal is to explore whether quantitative texture analysis can complement traditional US findings. Materials and Methods: 46 patients with 92 testis were examined using B-mode ultrasound and strain elastography (SE). A transverse US image was analyzed for each testis to extract histogram parameters such as mean, standard deviation, skewness, kurtosis, and entropy. SE was used to measure testicular elasticity. Results: Histogram analysis showed a significant correlation between age and homogenization of testicular tissue (r = 0.444; p

Background: Imaging modalities, such as ultrasonography (USG), can be used to evaluate and monitor the musculoskeletal system during the clinical progression of Duchenne muscular dystrophy (DMD). Aim: This study aimed to measure passive muscle stiffness in children with Duchenne muscular dystrophy and to compare these measurements with those of healthy children. Methods: Patients with DMD were evaluated clinically (age, clinical functional score, timed Gover score), serum creatine kinase level, B-mode ultrasonography, and real-time tissue elastography imaging. Results: A total of 64 boys were included in this study. The medial and lateral gastrocnemius muscle strain ratio in patients with DMD was significantly bigger than that in the control group (medial; 1.66 ± 1.23 vs 0.81 ± 0.16, P < 0.001, lateral; 1.49 ± 0.52 vs 0.85 ± 0.16, P < 0.001). Conclusion: The strain ratios of the medial and lateral gastrocnemius muscles were greater in patients with Duchenne muscular dystrophy than in the controls, indicating that ultrasound elastography may be beneficial for diagnosis and follow-up.