Journal of the Korean Society of Radiology最新文献

Deep learning-based segmentation has become a key tool for the precise and automated analysis of anatomical structures, such as bones, cartilage, and muscles, in musculoskeletal (MSK) imaging. This study examined the research trends by analyzing the number of related publications in PubMed since 2016 from both clinical and technical perspectives. Early studies primarily focused on the segmentation of major anatomical structures such as the spine and knee using large-scale datasets. However, recent studies have expanded to include the extremities and shoulders. In lesion segmentation, traditional topics such as body composition analysis, fractures, and tumors remain prominent, whereas deep learning-based detection and classification methods are increasingly integrated, leading to applications in newer areas. In addition, this study explored commonly used segmentation techniques and various applications of deep learning in MSK imaging. By systematically analyzing trends in deep learning-based segmentation research, we aim to provide insights into future directions for this rapidly evolving field.

Purpose: To evaluate whether early kinetic parameters from ultrafast dynamic contrast-enhanced (DCE) MRI correlate with the histopathological and morphological features of tumors.

Materials and methods: We retrospectively included 101 women with breast cancer (103 lesions; 85 invasive and 18 in situ) who underwent preoperative ultrafast DCE-MRI using compressed sensing between January 2020 and July 2022. Two radiologists assessed early kinetic parameters, time to enhancement (TTE) and maximum slope (MS), from the time-intensity curves. These were compared with prognostic factors.

Results: TTE was significantly shorter in tumors ≥2 cm (p = 0.001), invasive carcinomas (vs. carcinoma in situ; p = 0.003), estrogen receptor-negative tumors (p = 0.044), tumors with axillary lymph node metastasis (p = 0.021), and histological grade 3 tumors (vs. grade 1/2; p = 0.029). Higher MS was associated with tumors ≥2 cm (p = 0.008) and invasive carcinomas (p = 0.007). Multivariate regression confirmed tumor size ≥2 cm (p = 0.004 for TTE, p = 0.010 for MS) and invasive carcinoma (p = 0.012 for TTE, p = 0.015 for MS) as independent predictors of shorter TTE and higher MS. Inter-reader agreement for TTE and MS measurements was excellent, with intraclass correlation coefficient values of 0.951 and 0.879, respectively.

Conclusion: Early kinetic parameters from ultrafast DCE-MRI strongly correlated with certain clinicopathological prognostic factors of breast cancer.

Purpose: This study aimed to utilize functional MRI (fMRI) to assess neural responses to pre- and post-bimaxillary surgery self-images to establish an objective predictor of patient satisfaction with the surgical results.

Materials and methods: This prospective study included 16 patients who underwent bimaxillary surgery. Patients' fMRI data were obtained while viewing their self-images before and after surgery. Statistical analysis was performed using SPM12 software to identify changes in neural activation levels.

Results: Greater activations were recorded in the right middle frontal gyrus, right superior occipital gyrus/right superior parietal lobule, and right middle occipital gyrus in response to post-surgery images (false discovery rate-corrected p < 0.001). A positive correlation was found between patient satisfaction and the activation of the right superior temporal and precentral gyri when viewing pre- and post-surgery self-images, respectively.

Conclusion: Distinct neural activation patterns were observed when viewing preoperative and postoperative self-images, suggesting that fMRI could serve as a potential tool for predicting patient satisfaction with surgical outcomes. The activated brain regions may represent sites of esthetic satisfaction, aligning with patients' self-reported satisfaction levels.

In celebrating the 80th anniversary of the Korean Society of Radiology (KSR), this special report was prepared to reflect on the 60-year history of the Journal of the Korean Society of Radiology (JKSR) and to explore its future direction. This report is the first of a three-part series and summarizes the journal's evolution in various aspects, including its title changes, publication frequency, editorial structure, article types, the influence of advances in imaging equipment, and indexing in academic databases. In particular, as the JKSR increasingly expands its role as a society journal alongside its function as a scientific journal, in response to evolving demands, this report compares the JKSR's editorial direction and content with that of the Journal of the American College of Radiology. Based on this comparison, this report proposes strategic directions for the future development of the JKSR.

Purpose: To evaluate the feasibility of an abbreviated MRI protocol for distinguishing hepatic lesions during the initial staging of rectal cancer.

Materials and methods: We conducted a retrospective review of 255 abbreviated liver MRI (axial T2 weighted fat-suppressed, diffusion-weighted image [using three b-values: 50, 400, and 800], apparent diffusion coefficient) in the 3T unit. Two radiologists reviewed the images by consensus interpretation. We calculated 95% confidence intervals (CIs) for the diagnostic yield, prevalence, specificity, and sensitivity of abbreviated MRI for hepatic metastasis.

Results: Among the patients with too small to characterize (TSTC)-liver-on-CT, the specificity of abbreviated MRI for hepatic metastasis was 100% (20 of 20 patients; 95% CI: 0.8316-1.0000). Among patients suspected to have metastasis-liver-on-CT, the diagnostic yield of abbreviated MRI for hepatic metastasis was 90% (9 of 10 patients; 95% CI: 0.5550-0.9975), and the specificity of abbreviated MRI was 100% (1 of 1 patient; 95% CI: 0.0250-1.0000). None of the patients with TSTC-liver-on-CT-or suspected metastasis-liver-on-CT images showed unexpected hepatic metastasis at the 6-month follow-up CT.

Conclusion: Abbreviated liver MRI is useful for diagnosing the benignity of TSTC-liver-on-CT without full-protocol MRI in patients with newly diagnosed rectal cancer.

As the third report in celebration of the 80th anniversary of the Korean Society of Radiology, we aim to highlight the most viewed and most cited articles published in the Journal of the Korean Society of Radiology (JKSR) based on aggregated data. Additionally, we will review the recipients of the JKSR Outstanding Manuscript Award and the JKSR Distinguished Reviewer Award from the time these awards were established in 2016 through 2024. Through this report, we wish to express our sincere gratitude to all those who have contributed to the advancement of the JKSR.

Generative AI-including Generative Adversarial Networks, diffusion models, Large Language Models (LLMs), and more recently, vision-language models-is increasingly utilized in clinical practice for musculoskeletal imaging tasks such as disease diagnosis, image enhancement, image reconstruction, electronic health record summarization, and radiologic report generation. Integrating these technologies into radiology workflows can significantly advance radiology report generation, structured reporting, and patient-centered communication. However, challenges such as hallucination, bias, and performance drift remain persistent issues. Ensuring the safe and reliable use of LLMs in radiology requires domain-specific training, robust validation, and enhanced data privacy measures. This review summarizes available evidence regarding the potential utility of generative AI in musculoskeletal imaging and radiologic reporting, as well as the challenges and pitfalls in its application. Recommendations for future advancements and clinical translation are also discussed.