World journal of radiology最新文献

Large language models (LLMs) have emerged as transformative tools in radiology artificial intelligence (AI), offering significant capabilities in areas such as image report generation, clinical decision support, and workflow optimization. The first part of this manuscript presents a comprehensive overview of the current state of LLM applications in radiology, including their historical evolution, technical foundations, and practical uses. Despite notable advances, inherent architectural constraints, such as token-level sequential processing, limit their ability to perform deep abstract reasoning and holistic contextual understanding, which are critical for fine-grained diagnostic interpretation. We provide a critical perspective on current LLMs and discuss key challenges, including model reliability, bias, and explainability, highlighting the pressing need for novel approaches to advance radiology AI. Large concept models (LCMs) represent a nascent and promising paradigm in radiology AI, designed to transcend the limitations of token-level processing by utilizing higher-order conceptual representations and multimodal data integration. The second part of this manuscript introduces the foundational principles and theoretical framework of LCMs, highlighting their potential to facilitate enhanced semantic reasoning, long-range context synthesis, and improved clinical decision-making. Critically, the core of this section is the proposal of a novel theoretical framework for LCMs, formalized and extended from our group's foundational concept-based models - the world's earliest articulation of this paradigm for medical AI. This conceptual shift has since been externally validated and propelled by the recent publication of the LCM architectural proposal by Meta AI, providing a large-scale engineering blueprint for the future development of this technology. We also outline future research directions and the transformative implications of this emerging AI paradigm for radiologic practice, aiming to provide a blueprint for advancing toward human-like conceptual understanding in AI. While challenges persist, we are at the very beginning of a new era, and it is not unreasonable to hope that future advancements will overcome these hurdles, pushing the boundaries of AI in Radiology, far beyond even the most state-of-the-art models of today.

Thyroid-associated ophthalmopathy (TAO), an autoimmune disorder closely associated with thyroid dysfunction, requires timely diagnosis and ongoing accurate evaluation to improve patient outcomes. With the global incidence of TAO increasing and significantly affecting the quality of life of patients, there is an urgent need for effective diagnostic tools. As a noninvasive imaging technique, ultrasound plays a pivotal role in diagnosing and managing TAO, particularly in the early detection of and monitoring of disease progression. Despite its advantages, ultrasound faces challenges such as limited resolution for deep orbital structures and a lack of standardized protocols, which can lead to diagnostic inaccuracies. This paper reviews the current status of ultrasound applications in TAO, including diagnostic utility, recent technological advances, and key challenges. It proposes strategies for future research and improvement, emphasizing analysis of ultrasound imaging data to develop biomarker stratification models. We propose an integrated multimodal framework that combines ultrasound elastography with deep learning to improve diagnostic precision.

Imaging plays a crucial role in the evaluation of hepatocellular carcinoma (HCC) treatment response. Contrast enhanced computed tomography and magnetic resonance imaging with extra-cellular or hepatobiliary contrast agents are the imaging techniques of choice. Contrast enhanced ultrasound is a promising technique. In this paper, we describe radiological techniques, imaging findings after HCC treatment, and the criteria of response evaluation. The utility of the structured report is also evaluated.

Spontaneous intracerebral hemorrhage carries high early mortality and long-term disability, with hematoma expansion (HE) being the most important modifiable determinant of poor outcome. Although the computed tomography (CT) angiography (CTA) "spot sign" is a validated predictor of HE, it is not universally available, highlighting the need for accessible imaging tools. In this invited editorial, we discuss the study by Parry et al, who developed a simplified five-point prediction score based solely on non-contrast CT findings - baseline hematoma volume ≥ 30 mL, intraventricular hemorrhage, and the island, black hole, and swirl signs. Tested prospectively in 192 patients scanned within 4 hours of onset, the score showed a stepwise rise in HE risk from 7% at a score of 0% to 100% at a score of 5. We place these findings in the context of existing CTA and non-contrast CT literature and highlight their potential to accelerate triage and treatment, particularly where CTA is unavailable. Broader multicenter validation and integration with clinical and machine-learning approaches will further define the clinical impact of this streamlined, imaging-only tool.

Diffusion-weighted magnetic resonance imaging (DWI) has become an essential tool in the field of pancreatic magnetic resonance imaging, enabling the detection, characterization, prediction, and evaluation of pancreatic diseases. In this article, we review the acquisition parameters, postprocessing techniques, and quantitative methods utilized in pancreatic DWI. Various postprocessing models, including monoexponential, biexponential, stretched exponential and non-Gaussian kurtosis models, as well as deep learning networks, have been used to assess the clinical utility of these models in diagnosing pancreatic diseases. The single-shot echo-planar imaging sequence is the most commonly used sequence for DWI data acquisition in clinical settings, and the apparent diffusion coefficient (ADC) calculated using the monoexponential model is the most widely used quantitative parameter in clinical practice. The repeatability threshold for the ADC of a normal pancreas is 37% for test-retest scans, but the repeatability threshold for pancreatic tumors needs to be further investigated. Complex postprocessing models exploring novel DWI-based biomarkers beyond ADC to assess histological features, and artificial intelligence in DWI postprocessing and data analyses hold promise in the diagnosis of pancreatic diseases. Future work should focus on standardizing protocols, conducting multicentre studies, and exploring variety of methods to improve the accuracy of quantitative DWI results to increase the clinical effectiveness of DWI in patients with pancreatic diseases.

The application of artificial intelligence (AI) in carotid atherosclerotic plaque detection via computed tomography angiography (CTA) has significantly advanced over the past decade. This mini-review consolidates recent innovations in deep learning architectures, domain adaptation techniques, and automated plaque characterization methodologies. Hybrid models, such as residual U-Net-Pyramid Scene Parsing Network, exhibit a remarkable precision of 80.49% in plaque segmentation, outperforming radiologists in diagnostic efficiency by reducing analysis time from minutes to mere seconds. Domain-adaptive frameworks, such as Lesion Assessment through Tracklet Evaluation, demonstrate robust performance across heterogeneous imaging datasets, achieving an area under the curve (AUC) greater than 0.88. Furthermore, novel approaches integrating U-Net and Efficient-Net architectures, enhanced by Bayesian optimization, have achieved impressive correlation coefficients (0.89) for plaque quantification. AI-powered CTA also enables high-precision three-dimensional vascular segmentation, with a Dice coefficient of 0.9119, and offers superior cardiovascular risk stratification compared to traditional Agatston scoring, yielding AUC values of 0.816 vs 0.729 at a 15-year follow-up. These breakthroughs address key challenges in plaque motion analysis, with systolic retractive motion biomarkers successfully identifying 80% of vulnerable plaques. Looking ahead, future directions focus on enhancing the interpretability of AI models through explainable AI and leveraging federated learning to mitigate data heterogeneity. This mini-review underscores the transformative potential of AI in carotid plaque assessment, offering substantial implications for stroke prevention and personalized cerebrovascular management strategies.

Background: Obstructive sleep apnea-hypopnea syndrome (OSAHS) is a highly prevalent sleep-related respiratory disorder associated with serious health risks. Although polysomnography is the clinical gold standard for diagnosis, it is expensive, inconvenient, and unsuitable for population-level screening due to the need for professional scoring and overnight monitoring.

Aim: To address these limitations, this review aims to systematically analyze recent advances in deep learning-based OSAHS detection methods using snoring sounds, particularly focusing on graphical signal representations and network architectures.

Methods: A comprehensive literature search was conducted following the PRISMA 2009 guidelines, covering publications from 2010 to 2025. Studies were included based on predefined criteria involving the use of deep learning models on snoring sounds transformed into graphical representations such as spectrograms and scalograms. A total of 14 studies were selected for in-depth analysis.

Results: This review summarizes the types of signal modalities, datasets, feature extraction methods, and classification frameworks used in the current literatures. The strengths and limitations of different deep network architectures are evaluated.

Conclusion: Challenges such as dataset variability, generalizability, model interpretability, and deployment feasibility are also discussed. Future directions highlight the importance of explainable artificial intelligence and domain-adaptive learning for clinically viable OSAHS diagnostic tools.

Background: Spinal cord injury can lead to long-term disability, but current imaging methods are limited in predicting outcomes. Rapid diffusion tensor imaging (DTI) has shown promise, yet its clinical utility remains underexplored.

Aim: To evaluate the potential applications of a short DTI sequence, incorporated into a cervical spine magnetic resonance imaging (MRI) protocol, for characterizing a range of symptomatic spinal cord pathologies. We propose that cervical spine tractography can provide essential diagnostic information beyond what is currently available from conventional MRI.

Methods: We utilized a quick DTI sequence to create tractography models of the cervical spinal cord in four patients with distinct pathologies of various etiologies: Cord contusion, metastasis, myelopathy, and multiple sclerosis. We used DSI Studio software for post-processing of tractography cases. Fiber tract findings for each pathology case were compared to five control cases from the same scanner by looking for individual differences in white matter tract integrity based on the fractional anisotropy (FA) and mean diffusivity (MD) of the regions of interest from controls. These correlated with clinical presentations and conventional MRI findings.

Results: Control cases showed consistent and intact tract patterns with stable FA and MD values. In pathological cases, abnormalities in fiber orientation and tract continuity correlated with clinical symptoms and lesion locations.

Conclusion: The tractography models can provide additional information on white matter disruption that was not discernible on standard MRI sequences. However, its clinical use remains limited due to the need for specialized imaging protocols and complex post-processing, restricting its use to mostly academic settings.

Background: Pre-eclampsia is a significant challenge in obstetric care and adversely affects the feto-maternal outcomes, causing significant perinatal morbidity and mortality. Early detection of women at higher risk of developing pre-eclampsia in the first trimester provides a vital opportunity to initiate timely prophylactic therapies. First-trimester uterine artery Doppler is gaining prominence as a promising tool in early risk stratification.

Aim: To assess the role of uterine artery Doppler in screening for pre-eclampsia at 11-14 weeks of gestation.

Methods: Pregnant women attending routine antenatal care between 11 weeks and 14 weeks of gestation and undergoing first-trimester nuchal translucency screening were offered enrolment in the study. After calculating gestational age from the last menstrual period or fetal biometry (crown-rump length), Doppler ultrasound of bilateral uterine arteries was performed, and relevant Doppler parameters were recorded. Patients were followed until delivery for development of preeclampsia.

Results: Out of a total of 342 participants, 42 women (12.28%) developed preeclampsia, while the remaining 300 women (87.71%) had a normal pregnancy without preeclampsia. The mean uterine artery pulsatility index was significantly elevated in the pre-eclampsia group (1.9455 ± 0.36) compared to the normal group (1.474 ± 0.52) (P < 0.001). Using a pulsatility index threshold of 1.622, the receiver operating characteristic curve analysis demonstrated a sensitivity of 75% (95% confidence internal: 0.66-0.82), specificity of 86% (95% confidence internal: 0.78-0.91), positive predictive value of 84.27%, and negative predictive value of 77.48% with a diagnostic accuracy of 80.5%. The area under the curve was 0.896, indicating good diagnostic performance. Uterine artery notching was observed in 88% of the pre-eclampsia group compared to 16% in the control group, a difference that was statistically significant (P < 0.001).

Conclusion: Uterine artery Doppler in the first trimester at 11-14 weeks of gestation showed a good diagnostic value for forecasting the development of pre-eclampsia and holds promise as a valuable tool for early risk stratification.

Background: Thermal ablation (TA) has been proved to be effective and safe as minimally invasive treatment method for thyroid nodules. However, patients' experience during the procedures and quality of life varies among operators.

Aim: To explore strategy to improve quality of life and subjective experiences during TA for papillary thyroid carcinoma (PTC) based on thermal field management (TFM).

Methods: This retrospective propensity-matched cohort study was conducted in a single center. A total of 490 patients with PTC treated with TA from September 2023 to August 2024 were studied and divided into two groups (TFM group and non-TFM group) according to treatment strategies. Propensity score matching (PSM) was used to control for confounding factors. Complications, side effect and complaints of patients were compared between the two groups.

Results: A total of 113 patients (41.7 ± 10.6; 31 men, 82 women) were assigned to the TFM group, and 377 patients (mean age, 41.1 ± 10.7 year; 116 men, 261 women) were assigned to the non-TFM group. After PSM, a total of 108 patients were included in the TFM group, and 216 patients were included in the non-TFM group. The median follow-up was 10 months (range from 4-15 months). The incidence of voice change in the TFM group was significantly lower than that in the non-TFM group (0.9% vs 6.5%; P = 0.049). Although there was no statistically significant difference in rate of pain between the two groups, the proportion of complaining of pain in the TFM group was numerically lower than that in the non-TFM group (3.7% vs 9.7%, P = 0.090).

Conclusion: TFM, as a novel procedural optimization technique, can effectively improve quality of life and subjective experiences of patients during TA for PTC.