Frontiers in radiology最新文献

Spinal cord infarction following neuraxial anesthesia is a rare but serious complication. We present the case of a 70-year-old female who developed acute onset of left lower limb weakness immediately following spinal anesthesia administered for total hip replacement. Clinical features were consistent with incomplete Brown-Séquard syndrome. MRI revealed a T2/STIR hyperintense lesion involving the left hemicord at the D12-L1 vertebral level, suggestive of sulcal artery infarction. MRI showed only age-related changes. After a structured physiotherapy program, the patient experienced significant functional improvement and was discharged with stable vitals. This case highlights the importance of early diagnosis and management of spinal cord infarction in the perioperative setting.

Purpose: This study evaluates the impact of high-resolution T2-weighted imaging (T2_HR) combined with deep learning image reconstruction (DLR) on image quality, lesion delineation, and extraprostatic extension (EPE) assessment in prostate multiparametric MRI (mpMRI).

Materials and methods: This retrospective study included 69 patients who underwent mpMRI of the prostate on a 3 T scanner with DLR between April 2023 and March 2024. Routine mpMRI protocols adhering to the Prostate Imaging Reporting and Data System (PI-RADS) v2.1 were used, including an additional T2_HR sequence [2 mm slice thickness, 4:31 min vs. 4:12 min for standard T2 (T2_S)]. The image datasets were evaluated by two radiologists using a Likert scale ranging from 1 to 5, with 5 being the best for sharpness, lesion contours, motion artifacts, prostate border delineation, overall image quality, and diagnostic confidence. PI-RADS scoring and EPE suspicion were analyzed. The statistical methods used included the Wilcoxon signed-rank test and Cohen's kappa for inter-reader agreement.

Results: T2_HR significantly improved lesion contours (medians of 5 vs. 4, p < 0.001), prostate border delineation (medians of 5 vs. 4, p < 0.001), and overall image quality (medians of 5 vs. 4, p < 0.001) compared to T2_S. However, motion artifacts were significantly worse in T2_HR. Substantial inter-reader agreement was observed in the PI-RADS scoring. EPE detection marginally increased with T2_HR, though histopathological validation was limited.

Conclusion: T2_HR imaging with DLR enhances image quality, lesion delineation, and diagnostic confidence without significantly prolonged acquisition time. It shows potential for improving EPE assessment in prostate cancer but requires further validation in larger studies.

Objective: To systematically evaluate the diagnostic accuracy of various GPT models in radiology, focusing on differential diagnosis performance across textual and visual input modalities, model versions, and clinical contexts.

Methods: A systematic review and meta-analysis were conducted using PubMed and SCOPUS databases on March 24, 2025, retrieving 639 articles. Studies were eligible if they evaluated GPT model diagnostic accuracy on radiology cases. Non-radiology applications, fine-tuned/custom models, board-style multiple-choice questions, or studies lacking accuracy data were excluded. After screening, 28 studies were included. Risk of bias was assessed using the Newcastle-Ottawa Scale (NOS). Diagnostic accuracy was assessed as top diagnosis accuracy (correct diagnosis listed first) and differential accuracy (correct diagnosis listed anywhere). Statistical analysis involved Mann-Whitney U tests using study-level median (median) accuracy with interquartile ranges (IQR), and a generalized linear mixed-effects model (GLMM) to evaluate predictors influencing model performance.

Results: Analysis included 8,852 radiological cases across multiple radiology subspecialties. Differential accuracy varied significantly among GPT models, with newer models (GPT-4T: 72.00%, median 82.32%; GPT-4o: 57.23%, median 53.75%; GPT-4: 56.46%, median 56.65%) outperforming earlier versions (GPT-3.5: 37.87%, median 36.33%). Textual inputs demonstrated higher accuracy (GPT-4: 56.46%, median 58.23%) compared to visual inputs (GPT-4V: 42.32%, median 41.41%). The provision of clinical history was associated with improved diagnostic accuracy in the GLMM (OR = 1.27, p = .001), despite unadjusted medians showing lower performance when history was provided (61.74% vs. 52.28%). Private data (86.51%, median 94.00%) yielded higher accuracy than public data (47.62%, median 46.45%). Accuracy trends indicated improvement in newer models over time, while GPT-3.5's accuracy declined. GLMM results showed higher odds of accuracy for advanced models (OR = 1.84), and lower odds for visual inputs (OR = 0.29) and public datasets (OR = 0.34), while accuracy showed no significant trend over successive study years (p = 0.57). Egger's test found no significant publication bias, though considerable methodological heterogeneity was observed.

Conclusion: This meta-analysis highlights significant variability in GPT model performance influenced by input modality, data source, and model version. High methodological heterogeneity across studies emphasizes the need for standardized protocols in future research, and readers should interpret pooled estimates and medians with this variability in mind.

The increasing number of computed tomography (CT) scan examinations and the time-intensive nature of manual analysis necessitate efficient automated methods to assist radiologists in managing their increasing workload. While deep learning approaches primarily classify abnormalities from three-dimensional (3D) CT images, radiologists also incorporate clinical indications and patient demographics, such as age and sex, for diagnosis. This study aims to enhance multilabel abnormality classification and automated report generation by integrating imaging and non-imaging data. We propose a multimodal deep learning model that combines 3D chest CT scans, clinical information reports, patient age, and sex to improve diagnostic accuracy. Our method extracts visual features from 3D volumes using a visual encoder, textual features from clinical indications via a pretrained language model, and demographic features through a lightweight feedforward neural network. These extracted features are projected into a shared representation space, concatenated, and processed by a projection head to predict abnormalities. For the multilabel classification task, incorporating clinical indications and patient demographics into an existing visual encoder, called CT-Net, improves the F1 score to 51.58, representing a $+\Delta 6.13\%$ increase over CT-Net alone. For the automated report generation task, we extend two existing methods, CT2Rep and CT-AGRG, by integrating clinical indications and demographic data. This integration enhances Clinical Efficacy metrics, yielding an F1 score improvement of $+\Delta 14.78\%$ for the CT2Rep extension and $+\Delta 6.69\%$ for the CT-AGRG extension. Our findings suggest that incorporating patient demographics and clinical information into deep learning frameworks can significantly improve automated CT scan analysis. This approach has the potential to enhance radiological workflows and facilitate more comprehensive and accurate abnormality detection in clinical practice.

Objective: To evaluate the image quality and diagnostic efficacy of proton density-weighted MRI with intelligent quick magnetic resonance (iQMR) technology in the ankle joint injury.

Materials and methods: Forty-six patients with ankle injuries were prospectively enrolled, and proton density-weighted fat suppression imaging was performed on a 3.0T MRI scanner using both an iQMR protocol (48.28 s) and a Conventional protocol (113.00 s), respectively. The original image was processed using iQMR to improve spatial resolution and reduce noise interference. Thus, four sets of images (iQMR raw, iQMR-processed, Conventional raw, and Conventional-processed) were generated. Image quality and diagnostic efficacy were assessed by objective metrics (signal-to-noise ratio, SNR and contrast-to-noise ratio, CNR), subjective scores (tissue edge clarity/sharpness, signal uniformity, fat suppression uniformity, vascular pulsation artifacts, and overall image quality), and ligaments/tendons injury grade.

Results: The SNRs (tibia, talus, etc.) and CNRs (talus-flexor hallucis longus, etc.) of iQMR-processed images were significantly higher than those of Conventional raw images (P < 0.05), except for the SNR of Achilles tendon (P > 0.05). And the iQMR-processed images were superior to the Conventional raw images in the scores of edge clarity/sharpness, signal uniformity and overall image quality (P < 0.05), with no significant differences in fat suppression uniformity and vascular pulsation artifacts (P > 0.05). There was no significant difference among the four groups of images in ligaments/tendons injury grading (P > 0.05), but the iQMR-processed images improved diagnostic confidence [κ (kappa) = 0.919].

Conclusion: The iQMR technology can effectively shorten the scan time, improve the image quality without affecting the diagnostic accuracy, which is especially suitable for the motion artifacts-sensitive patients and optimizes clinical workflow.

Cerebellar microhemorrhages have not been previously documented in methylmalonic acidemia with homocystinuria (MMA-HC), a rare inherited metabolic disorder. Herein, we reported an 18-year-old female presented with acute gait instability and dysarthria post-febrile illness. Biochemical testing revealed severe hyperhomocysteinemia. Brain MRI demonstrated bilateral cerebellar DWI/FLAIR hyperintensities. Whole-exome sequencing confirmed compound heterozygous MMACHC mutations, establishing cblC-type MMA-HC diagnosis. Symptoms resolved after one month of vitamin-based therapy. Follow-up 3.0 T MRI and 7.0 T MRI susceptibility-weighted imaging (SWI) uncovered multiple punctate cerebellar vermian microhemorrhages-a previously unreported finding. This case highlights an unusual adult-onset presentation of MMA-HC and represents the first report of SWI-detectable cerebellar vermis microhemorrhages with this condition, visualized. This finding suggests that cerebellar microhemorrhages may be an under-recognized feature in MMA-HC, particularly detectable using high-field SWI during acute exacerbations, and contributes to a more comprehensive understanding of the neurological complications in this metabolic disorder.

Introduction: This study investigates the application of a deep learning model, YOLOv8-Seg, for the automated classification of osteoporotic vertebral fractures (OVFs) from computed tomography (CT) images.

Methods: A dataset of 673 CT images from patients admitted between March 2013 and May 2023 was collected and classified according to the European Vertebral Osteoporosis Study Group (EVOSG) system. Of these, 643 images were used for training and validation, while a separate set of 30 images was reserved for testing.

Results: The model achieved a mean Average Precision (mAP50-95) of 85.9% in classifying fractures into crush, anterior wedge, and biconcave types.

Discussion: The results demonstrate the high proficiency of the YOLOv8-Seg model in identifying OVFs, indicating its potential as a decision-support tool to streamline the current manual diagnostic process. This work underscores the significant potential of deep learning to assist medical professionals in achieving early and precise diagnoses, thereby improving patient outcomes.

Background: Large language models (LLMs) appear to be capable of performing a variety of tasks, including answering questions, but there are few studies evaluating them in direct comparison with clinicians. This study aims to compare the performance of artificial intelligence (AI) models and clinical specialists in informing patients about varicocele embolization. Additionally, we aim to establish an evidence base for future hybrid informational systems that integrate both AI and clinical expertise.

Methods: In this prospective, double-blind, randomized controlled trial, 25 frequently asked questions about varicocele embolization (collected via Google Search trends, patient forums, and clinical experience) were answered by three AI models (ChatGPT-4o, Gemini Pro, and Microsoft Copilot) and one interventional radiologist. Responses were randomized and evaluated by two independent interventional radiologists using a valid 5-point Likert scale for academic accuracy and empathy.

Results: Gemini achieved the highest mean scores for both academic accuracy (4.09 ± 0.50, 95% CI: 3.95-4.23) and higher expert-rated scores for empathetic communication (3.54 ± 0.59, 95% CI: 3.38-3.70), followed by Copilot (academic: 4.07 ± 0.46, 95% CI: 3.94-4.20; empathy: 3.48 ± 0.53, 95% CI: 3.33-3.63), ChatGPT (academic: 3.83 ± 0.58, 95% CI: 3.67-3.99; empathy: 2.92 ± 0.78, 95% CI: 2.70-3.14), and the comparator physician (academic: 3.75 ± 0.41, 95% CI: 3.64-3.86; empathy: 3.12 ± 0.82, 95% CI: 2.89-3.35). ANOVA revealed statistically significant differences across groups for both academic accuracy (F = 6.181, p < 0.001, η ² = 0.086) and empathy (F = 9.106, p < 0.001, η ² = 0.122). Effect sizes were medium for academic accuracy and large for empathy.

Conclusions: AI models, particularly Gemini, received higher ratings from expert evaluators compared to the comparator physician in patient education regarding varicocele embolization, excelling in both academic accuracy and empathetic communication style. These preliminary findings suggest that AI models hold significant potential to complement patient education systems in interventional radiology practice and provide compelling evidence for the development of hybrid patient education models.

Background: According to the 5th revision of World Health Organization (WHO) of central nervous system tumors classification, gliosarcoma is a malignant tumor grade 4 and is the rarest and aggressive subtype of isocitrate dehydrogenase (IDH) wild-type glioblastoma. The special histopathological feature of the tumor is its biphasic differentiation including both the glial and the sarcomatous (mesenchymal) components of the tumor. The characteristics mentioned above create difficulties in radiological and histological diagnoses. Because of its rarity, gliosarcoma is typically not even considered in the differential diagnosis.

Case presentation: This clinical case study describes a 55-year-old man exhibiting acute right-sided hemiparesis and disorientation for 12 h with loss of consciousness. A brain МRI of the patient revealed an intracerebral mass in the left frontoparietal area with close relationship with the dura mater, ring-like enhancement, severe perifocal edema, restricted diffusion of the solid component, internal vascular shunts, microhemorrhages, and elevated perfusion values. At the preoperative stage, the differential diagnosis included glioblastoma, solitary metastasis, and the possibility of an anaplastic meningioma. Tumor microsurgical resection was performed. According to the results of histological and immunohistochemical studies, gliosarcoma was diagnosed.

Discussion: The only characteristic gliosarcoma feature was the phenomenon of solid node heterogeneity detected on the conventional T2-weighted sequence: a combination of hypo- and hyperintense parts. While multiparametric magnetic resonance imaging (MRI) aids in differentiating high-grade gliomas, metastases, and meningiomas, gliosarcoma remains underrecognized because of overlapping features. The observed T2 heterogeneity may serve as a potential radiological marker for gliosarcoma. Accurate and timely identification of brain tumor type is required to establish the appropriate extent of resection in surgical planning.

Conclusion: This case publication does not intend to ignore the data of conventional sequences and instead considers them to be included in the structure of the multiparametric MRI protocol. However, larger studies are needed to validate the findings of this case study and refine diagnostic criteria for this rare tumor.