Korean Journal of Radiology最新文献

Multimodal large language models (MLLMs) are emerging as powerful tools in medicine, particularly in radiology, with the potential to serve as trusted artificial intelligence (AI) partners for clinicians. In radiology, these models integrate large language models (LLMs) with diverse multimodal data sources by combining clinical information and text with radiologic images of various modalities, ranging from 2D chest X-rays to 3D CT/MRI. Methods for achieving this multimodal integration are rapidly evolving, and the high performance of freely available LLMs may further accelerate MLLM development. Current applications of MLLMs now span automatic generation of preliminary radiology report, visual question answering, and interactive diagnostic support. Despite these promising capabilities, several significant challenges hinder widespread clinical adoption. MLLMs require access to large-scale, high-quality multimodal datasets, which are scarce in the medical domain. Risks of hallucinated findings, lack of transparency in decision-making processes, and high computational demands further complicate implementation. This review summarizes the current capabilities and limitations of MLLMs in medicine-particularly in radiology-and outlines key directions for future research. Critical areas include incorporating region-grounded reasoning to link model outputs to specific image regions, developing robust foundation models pre-trained on large-scale medical datasets, and establishing strategies for the safe and effective integration of MLLMs into clinical practice.

Antibody-drug conjugates (ADCs) have revolutionized the treatment landscape for advanced bladder cancer, particularly enfortumab vedotin and trastuzumab deruxtecan, which target Nectin-4 and human epidermal growth factor receptor 2 (HER 2), respectively. These ADCs have shown substantial efficacy, improving survival in patients who have progressed after chemotherapy and immunotherapy. Imaging plays a pivotal role in ADC-based therapy, extending beyond diagnosis and staging to assessing treatment response, detecting recurrence, and evaluating toxicity. Computed tomography (CT), multiparametric magnetic resonance imaging (MRI), and fluorodeoxyglucose positron emission tomography (FDG-PET) are widely used for these purposes. Despite the efficacy of ADCs, resistance mechanisms such as antigen loss and payload resistance continue to pose challenges, necessitating the development of next-generation ADCs. Response assessment largely relies on Response Evaluation Criteria in Solid Tumors (RECIST) 1.1, with growing interest in multiparametric MRI for evaluating complete response in bladder lesions. Additionally, imaging helps identify ADC-related toxicities, including pneumonitis and gastrointestinal complications. Radiologists must be aware of these evolving therapeutic and imaging paradigms to optimize patient management. The integration of imaging with ADC-based treatment requires a multidisciplinary approach to improve outcomes. This review highlights the critical role of imaging in ADC therapy and underscores the need for radiologists to adapt to these advancements in bladder cancer treatment.

Cardiac sarcoidosis (CS) poses significant diagnostic and therapeutic challenges due to its heterogeneous clinical manifestations and the limitations of conventional diagnostic approaches. Advances in imaging modalities, particularly cardiac magnetic resonance imaging (CMR) and ¹⁸F-fluorodeoxyglucose positron emission tomography (FDG-PET), have revolutionized the evaluation and management of this complex condition. CMR, with its superior spatial resolution and advanced techniques such as late gadolinium enhancement, T1/T2 mapping, and extracellular volume quantification, offers unparalleled insights into myocardial structure and fibrosis. These techniques not only enhance diagnostic accuracy but also provide critical information on disease activity and treatment response. Among these, T2 mapping has emerged as a valuable marker for active inflammation, with high values reliably indicating acute disease states. FDG-PET serves as a complementary modality by detecting active granulomatous inflammation and guiding immunosuppressive therapy. The synergistic integration of CMR and FDG-PET provides a comprehensive approach to diagnosing and monitoring CS, enabling the identification of subclinical disease and the optimization of therapeutic strategies. Furthermore, the incorporation of quantitative biomarkers, such as strain metrics and T2 values, promises to refine disease assessment and management. These advancements have the potential to transform the paradigm of CS care, ultimately improving patient outcomes.

Objective: To evaluate the accuracy of large language models (LLMs) in extracting Coronary Artery Disease-Reporting and Data System (CAD-RADS) 2.0 components from coronary CT angiography (CCTA) reports, and assess the impact of prompting strategies.

Materials and methods: In this multi-institutional study, we collected 319 synthetic, semi-structured CCTA reports from six institutions to protect patient privacy while maintaining clinical relevance. The dataset included 150 reports from a primary institution (100 for instruction development and 50 for internal testing) and 169 reports from five external institutions for external testing. Board-certified radiologists established reference standards following the CAD-RADS 2.0 guidelines for all three components: stenosis severity, plaque burden, and modifiers. Six LLMs (GPT-4, GPT-4o, Claude-3.5-Sonnet, o1-mini, Gemini-1.5-Pro, and DeepSeek-R1-Distill-Qwen-14B) were evaluated using an optimized instruction with prompting strategies, including zero-shot or few-shot with or without chain-of-thought (CoT) prompting. The accuracy was assessed and compared using McNemar's test.

Results: LLMs demonstrated robust accuracy across all CAD-RADS 2.0 components. Peak stenosis severity accuracies reached 0.980 (48/49, Claude-3.5-Sonnet and o1-mini) in internal testing and 0.946 (158/167, GPT-4o and o1-mini) in external testing. Plaque burden extraction showed exceptional accuracy, with multiple models achieving perfect accuracy (43/43) in internal testing and 0.993 (137/138, GPT-4o, and o1-mini) in external testing. Modifier detection demonstrated consistently high accuracy (≥0.990) across most models. One open-source model, DeepSeek-R1-Distill-Qwen-14B, showed a relatively low accuracy for stenosis severity: 0.898 (44/49, internal) and 0.820 (137/167, external). CoT prompting significantly enhanced the accuracy of several models, with GPT-4 showing the most substantial improvements: stenosis severity accuracy increased by 0.192 (P < 0.001) and plaque burden accuracy by 0.152 (P < 0.001) in external testing.

Conclusion: LLMs demonstrated high accuracy in automated extraction of CAD-RADS 2.0 components from semi-structured CCTA reports, particularly when used with CoT prompting.

Objective: In this study, we investigated whether artificial intelligence (AI) analysis of chest radiographs (CXRs) can predict major adverse clinical events in patients visiting the emergency department (ED) with acute cardiopulmonary symptoms.

Materials and methods: This secondary analysis of a previous clinical trial included patients who visited the ED with symptoms suggestive of acute cardiopulmonary disease and underwent chest radiography between June 2020 and December 2021. All patients underwent triage upon arrival at ED according to the Korean Triage and Acuity Scale (KTAS). The CXRs were retrospectively analyzed using a commercial AI (Lunit INSIGHT CXR, version 3.1.4.1) capable of detecting seven abnormalities on a single frontal CXR. The predictive performance of the AI analysis for major adverse cardiopulmonary events (any among hospitalization, ED revisits, and death in the ED due to acute cardiopulmonary disease) was compared with that of the KTAS using the area under the receiver operating characteristic curve (AUC). Multivariable (the AI analysis result and KTAS level) logistic regression analysis was conducted to investigate whether the AI analysis result was an independent predictor of the events and whether the combination of the AI analysis and KTAS has additional merit.

Results: Among 3576 patients (1966 males; mean age, 64 years), 1148 (32.1%) experienced major adverse cardiopulmonary events. AI analysis of CXRs outperformed the KTAS in predicting these events (AUC, 0.795 vs. 0.610; P < 0.001). The AI analysis result was an independent predictor of these events after adjusting for the KTAS level (adjusted odd ratios of 1.032 and 6.913 for every 1% increase and ≥15%, respectively, in the AI probability score; P < 0.001). The combination of the AI analysis and KTAS showed an AUC that was higher than that of the KTAS alone (0.799; P < 0.001) and in-par with that of the AI analysis only (P = 0.187).

Conclusion: AI analysis of CXRs showed greater accuracy than the KTAS did in predicting major adverse cardiopulmonary events in patients visiting the ED with acute cardiopulmonary symptoms. AI analysis may enhance the efficacy of patient triage in the ED.

Objective: Plasmacytoid urothelial carcinoma (PUC) is a rare aggressive bladder cancer subtype with limited imaging data owing to its low incidence. This study aimed to report the characteristic features of PUC on multiparametric MRI (mpMRI).

Materials and methods: We retrospectively analyzed 13 patients with histologically confirmed PUC who underwent preoperative mpMRI between January 2019 and August 2024. Two blinded radiologists independently assessed tumor size, morphology, signal intensity, apparent diffusion coefficient (ADC) values, dynamic contrast enhancement patterns, contrast enhancement features, and invasive characteristics. Vesical imaging-reporting and data system (VI-RADS) scores were recorded. Interobserver agreement was evaluated using the kappa statistic.

Results: PUC predominantly exhibited diffuse (6/13, 46.2%) or localized (5/13, 38.5%) bladder wall thickening. Diffuse thickening was often associated with a linitis plastica-like appearance. On high b-value diffusion-weighted imaging (DWI), eight and seven cases depending on readers (61.5% and 53.8%, respectively) showed mild hyperintensity or isointensity, with a mean ADC value of 1.1 × 10⁻³ mm²/s. Dynamic contrast-enhanced MRI revealed progressive and prolonged enhancement in 10 cases (76.9%). VI-RADS scores ≥ 4 were observed in 11 cases (84.6%). Histopathological analysis showed that tumors with progressive and prolonged enhancement contained myxoid stroma and some fibrous tissue. Interobserver agreement was excellent for most imaging features, except for good agreement on DWI signal intensity.

Conclusion: PUC demonstrates notable mpMRI features, including localized or diffuse wall thickening (often with a linitis plastica-like appearance), muscle-invasive and advanced disease, progressive and prolonged enhancement patterns, and mild hyperintensity or isointensity on high b-value DWI. These features, which are potentially linked to the myxoid stromal composition of the tumor, suggest that mpMRI may serve as a noninvasive diagnostic tool for this aggressive malignancy. However, further studies with larger cohorts are required to confirm these findings.