European Radiology最新文献

Objectives: Prior publications have highlighted inconsistent labelling of intended use cases and target population for U.S. Food and Drug Administration-approved AI medical devices, especially for children. The extent to which this issue applies to devices within Europe remains unaddressed.

Materials and methods: A comprehensive review was conducted of all regulatory-approved AI medical devices for use in radiology from a non-profit, publicly available database. Two independent reviewers assessed information about all the devices regarding use case, modality and intended population. A third reviewer resolved any discrepancies. Where the intended population was unclear, a standardised review of the available evidence and marketing materials for the AI device was conducted.

Results: Only four (4/213, 2%) AI medical devices were clearly labelled for paediatric use. A further 11% were intended for all ages, including children. The majority (88/213; 41%) of all AI medical devices did not clearly demonstrate their intended population on the database. Further examination of the scientific literature and marketing of these "unclear" devices showed that 6 (6/88, 7%) of these included patients under 18 in their intended target population, but 47% (41/88) still remained unclear after further review.

Conclusion: Most regulated radiology AI medical devices have missing or unclear information regarding the appropriate use in children. This poses significant potential risk, including inadvertent off-label use, which could compromise patient safety. The EU AI Act emphasises the need for transparency and accountability in AI device deployment, and we therefore advocate for a standardised paediatric safety indicator to clearly communicate suitability.

Key points: Question To what extent do European-approved radiology AI devices provide clear labelling of intended use and suitability for paediatric populations? Findings Many radiology AI medical devices in Europe lack explicit paediatric use information, raising concerns about unintended off-label use. Clinical relevance Clear labelling of AI device suitability for children is essential to ensure safe use. A standardised safety indicator could aid clinicians in appropriate device selection.

Objectives: To evaluate the relationship between mucin quantity on MRI and outcomes following total neoadjuvant therapy (TNT) in patients with rectal cancer.

Materials and methods: This retrospective, single-center study included patients with rectal adenocarcinoma who underwent TNT followed by surgery or non-operative management (NOM) from January 2018-December 2019. Two abdominal radiologists independently scored baseline and restaging MRIs for mucin quantity in the tumor or tumor bed, using two classifications: < 50% vs. ≥ 50% (similar to histological classification) and no mucin vs. any mucin. Statistical analysis was performed to determine if patient outcomes differed between tumors of different mucin quantities.

Results: Among 189 patients, 114/189 (60%) male, 75/189 (40%) female), the median age at diagnosis was 57 years (IQR: 48-68). On baseline MRI, 29/189 (15%) patients had tumor containing any amount of mucin, and on restaging MRI, 27/189 (14%) patients had tumor containing any mucin. 103/189 (54%) underwent surgery after TNT, and 86/189 (46%) underwent NOM. 59/189 (31%) experienced local recurrence, distant recurrence, or local regrowth. There were no significant differences in the rate of pathologic complete response/sustained clinical complete response, time to local recurrence/local regrowth, time to distant recurrence, and disease-free survival between tumors of different mucin quantity, regardless of the timing and criteria for determining mucin quantity (< 50% vs. ≥ 50% mucin criteria, and no vs. any mucin criteria on baseline MRI, restaging MRI and on histopathology, respectively).

Conclusion: Mucin on restaging MRI should not rule out NOM or prompt more aggressive therapy in the absence of other high-risk features.

Key points: Question Is there a correlation in mucin content of tumor and patient outcomes in the setting of total neoadjuvant therapy followed by surgery or nonoperative management? Finding There were no prognostic differences between tumors of different mucin quantity in multiple outcomes, including local recurrence, local regrowth, recurrence, and disease-free survival. Clinical relevance Although prior studies have shown conflicting results, our findings suggest mucin content on MRI may not predict worse outcomes in rectal cancer patients treated with contemporary total neoadjuvant therapy, including those managed nonoperatively.

Objectives: The aim of this study was to describe the imaging features on dynamic CT and MRI of a series of pathologically confirmed low-grade vascular neoplasia of the liver (LGVNL).

Materials and methods: In this retrospective multicenter study, patients diagnosed with pathologically proven LGVNL between January 2014 and August 2024 and with cross-sectional imaging (CT or MRI) were included. Based on prior studies, we divided the patients into two groups: a group with typical LGVNL features and a group with atypical tumors. Univariable analysis and the logistic regression model were used to evaluate the outcome of typical and atypical LGVNL features.

Results: Twenty-eight patients were included (20 men, mean age 53.7 ± 13.4 [SD] years old). The median size of tumors at diagnosis was 22 mm [IQR, 10-80]. A typical LGVNL pattern including thick continuous peripheral arterial phase "flower petal shape" enhancement was found on MRI in 67% (18/27) with lobulated lesions (p = 0.001), a marked hypersignal on T2-weighted images (p = 0.003) and a high apparent diffusion coefficient (ADC) (mean tumor ADC, 2.1 × 10^-3 ± 0.3 [SD] mm²/s) (p = 0.006) with portal and delayed phase filling following contrast administration. An atypical LGVNL pattern was found in 33% (9/27), including homogeneous arterial phase enhancement that persisted during the portal and delayed phases. After a median follow-up of 16 months, tumor growth was observed in 42% (8/19) and was faster in patients with the atypical LGVNL pattern (p = 0.013).

Conclusion: The typical imaging pattern of LGVNL that includes arterial phase "flower petal shape" enhancement with a marked hypersignal on T2-weighted images and a high ADC was seen in most cases.

Key points: Question Low-grade vascular neoplasia of the liver, including hepatic small vessel neoplasms and anastomosing hemangioma, is a new entity of vascular tumors with few radiological descriptions. Findings Diagnostic imaging criteria include the "flower petal shape" enhancement pattern on arterial phase, a marked hypersignal on T2-weighted images and high ADC values. Clinical relevance This multicenter study provides clinical, pathological and imaging data on low-grade vascular neoplasia of the liver and highlights specific imaging features for its diagnosis. Knowledge of these imaging features helps eliminate differential diagnoses.

Objectives: Computed tomography (CT) is widely used for bone health assessment, impacting osteoporosis diagnosis and treatment. However, the influence of intravenous contrast agents on CT-based bone mineral density (BMD) measurements remains debated. This study evaluates the effect of contrast agents on Hounsfield measurements, T-scores, and Z-scores, assessing their impact on diagnostic accuracy to reduce misclassification and optimize CT-based BMD assessment.

Materials and methods: A retrospective analysis of 597 patients (median age: 66 years, 157 females, 440 males) was performed using dual-energy CT (DECT) scans of the abdomen and chest. All patients underwent non-contrast, arterial, and venous phase CT. Automated segmentation (nnU-Net) delineated L1 and L1-L4 trabecular bone, validated by two radiologists. T-scores were calculated according to DEXA-equivalent guidelines.

Results: Based on non-contrast CT, 35% were diagnosed with osteoporosis, 46% with osteopenia, and 18% had normal bone status. Median T-score was -2.0 (L1) and -2.1 (L1-L4) (p < 0.001). Contrast agents significantly altered BMD values, with median changes of 22.9% (arterial) and 20.1% (venous). The most pronounced changes occurred in patients under 50 years (+99% at L1, p < 0.001). In older females, 21% were misclassified as osteopenic instead of osteoporotic (p < 0.001).

Conclusions: Contrast agents significantly affect BMD measurements, leading to diagnostic misclassification. This effect should be considered when using CT for osteoporosis diagnosis and treatment planning.

Key points: Question Standard CT scans with contrast media may distort bone density measurements, potentially leading to misdiagnosis of osteoporosis and inappropriate clinical decisions. Findings Contrast-enhanced CT scans significantly alter T- and Z-scores, leading to diagnostic shifts in over 50% of patients, especially women over 50. Clinical relevance Our findings highlight the risk of osteoporosis misclassification due to contrast agents in CT imaging, underscoring the need for adjusted interpretation protocols to ensure accurate diagnosis and appropriate treatment, particularly in older adults and female patients.

Objectives: The rising global cancer burden underscores the need for efficient screening strategies. Whole-body magnetic resonance imaging (WB-MRI) has emerged as a promising modality for cancer screening, with growing use in research and commercial settings. This study aimed to evaluate the opportunistic cancer detection rate and the feasibility of WB-MRI in asymptomatic individuals.

Materials and methods: A systematic review and meta-analysis were conducted per PRISMA guidelines. A literature search was performed across multiple databases from January 2015 to April 2025. Eligible studies used WB-MRI for cancer detection in asymptomatic individuals. Studies were excluded if they combined WB-MRI with other imaging methods or included patients with active malignancy or comorbidities/genetic syndromes associated with increased cancer risk. Random-effects meta-analyses estimated pooled proportions of confirmed cancer diagnoses. Risk of bias was assessed using the ROBINS-I tool. Sensitivity and subgroup analyses, publication bias assessment, and meta-regression were performed.

Results: Ten studies were included, comprising 9024 participants. The pooled detection rate for confirmed cancer was 1.57% (95% CI: 1.22-2.03%; I² = 31.3%). Results were robust in sensitivity and meta-regression analyses. No significant subgroup differences or publication bias were found. Most studies had a moderate to serious risk of bias.

Conclusion: Although WB-MRI shows potential as an opportunistic non-invasive cancer detection tool, modest detection rates, frequent incidental findings, unstandardized protocols, and lack of long-term outcome or cost-effectiveness data limit its current clinical utility.

Key points: Question Can whole-body magnetic resonance imaging (WB-MRI) serve as an effective cancer detection tool for asymptomatic individuals across various clinical and geographic settings? Findings This meta-analysis of over 9000 asymptomatic individuals found a lack of information on cost-effectiveness, unstandardized protocols, a modest cancer detection rate and high rates of incidental findings. Clinical relevance Despite the need for effective cancer screening tools and growing popularity in commercial and research centers, whole-body MRI lacks sufficient diagnostic yield, follow-up reports and standardization for opportunistic cancer detection in asymptomatic individuals and may lead to unnecessary investigations.

Objectives: The aim of this study was to compare in vivo and ex vivo Photon Counting CT (PCCT) of subchondral bone features in patients with knee osteoarthritis (KOA).

Materials and methods: Pre-surgery in vivo and post-surgery ex vivo PCCT of the tibial plateau from participants with severe KOA referred to arthroplasty surgery from January 2022 through September 2023 were compared. Linear regression and Bland-Altman plots were used to assess correlation and agreement between in vivo and ex vivo measures of bone volume fraction (BV/TV), trabecular thickness (Tb.Th.) and attenuation in healthy and sclerotic trabecular bone. Delineated areas of bone sclerosis were compared using the Dice coefficient and Hausdorff distance.

Results: 18 in vivo/ex vivo PCCT scans were included. Strong correlations were found for BV/TV, R² = 0.82 and attenuation; healthy, R² = 0.89, and sclerotic, R² = 0.79, bone, while a moderate correlation was found for Tb.Th., R² = 0.55. Bias for BV/TV and Tb.Th. was -4.1% and -0.598 mm, respectively, and -41.4 HU and -81.1 HU for healthy and sclerotic bone, respectively. A proportional bias was observed for Tb.Th. and BV/TV. There was excellent agreement between the segmentations of sclerotic areas (Dice coefficient = 0.91, Hausdorff distance = 0.11 mm).

Conclusion: In patients with severe KOA, BV/TV and attenuation can be obtained with a high correlation and small bias between in vivo and ex vivo scans, while Tb.Th. showed moderate correlation and larger bias. Longitudinal studies using in vivo PCCT are feasible, but caution may be advised when measuring Tb.Th. The key OA feature of subchondral bone sclerosis is well translated from ex vivo to in vivo PCCT.

Key points: Question Bone changes occur with osteoarthritis development; the role of these changes is unclear, and no method for visualising bone microstructure in vivo exists. Findings Photon-counting CT showed a strong correlation between in vivo and ex vivo subchondral density measures, while a moderate correlation was found for trabecular thickness. Clinical relevance Photon-counting CT is feasible for in vivo longitudinal evaluation of bone in patients with knee osteoarthritis, allowing studies into the earlier stages of the disease.

Low-dose CT screening for lung cancer reduces the risk of death from lung cancer by at least 21% in high-risk participants and should be offered to people aged between 50 and 75 with at least 20 pack-years of smoking. Iterative reconstruction or deep learning algorithms should be used to keep the effective dose below 1 mSv. Deep learning algorithms are required to facilitate the detection of nodules and the measurement of their volumetric growth. Only large solid nodules larger than 500 mm³ or those with spiculations, bubble-like lucencies, or pleural indentation and complex cysts should be investigated further. Short-term follow-up at 3 or 6 months is required for solid nodules of 100 to 500 mm³. A watchful waiting approach is recommended for most subsolid nodules, to limit the risk of overtreatment. Finally, the description of additional findings must be limited if LCS is to be cost-effective. KEY POINTS: Low-dose CT screening reduces the risk of death from lung cancer by at least 21% in high-risk individuals, with a greater benefit in women. Quality assurance of screening is essential to control radiation dose and the number of false positives. Screening with low-dose CT scans detects incidental findings of variable clinical relevance, only those of importance should be reported.

Objectives: Reducing motion artifacts in coronary computed tomography angiography (CCTA) is essential for accurate coronary artery disease assessment. We evaluated the clinical performance of a motion-compensated reconstruction (MCR) using subjective image quality (SIQ) and interpretability of CCTA at varying heart rates (HR).

Materials and methods: We retrospectively identified 150 patients, grouped by HR (≤ 60, 60-69, ≥ 70 bpm, n = 50 each), referred for prospective ECG-gated CCTA on a spectral dual-layer CT. Two blinded observers independently assessed SIQ on a per-segment (≥ 1.5 mm) and per-patient level using a 4-point Likert scale in 18 coronary segments (78% RR-interval). Sufficient diagnostic interpretability was defined as SIQ ≥ 2. Per-vessel scores were calculated excluding side branch segments. Per-segment SIQ interobserver agreement was assessed using Cohen's Weighted Kappa. Between MCR and standard reconstruction (SR) at 78% RR-interval, SIQ was compared with Wilcoxon signed-rank tests and diagnostic interpretability and HR-categories using McNemar tests.

Results: Mean age was 57 (50-64) years, with 50% men, and 1970 included segments. Interobserver agreement was 0.80 for SR and 0.77 for MCR. Positive trends of improved SIQ were seen across all HR-categories and levels, with significant improvements in all but ≥ 70 bpm on a patient level (p = 0.22). Likewise, positive trends were seen for diagnostic interpretability across all levels and HR-categories with significant improvements at the per-segment level for HR > 60 bpm and per-patient level for 61-69 bpm.

Conclusion: Compared to the standard reconstruction at 78% RR-interval, MCR significantly improves SIQ and diagnostic interpretability in patients referred for CCTA in most HRs and major vessels (≥ 1.5 mm).

Key points: Question Motion artifacts hinder the assessment of coronary arteries on coronary CT angiography (CCTA), leading to more non-diagnostic segments or scans. Findings Compared to the standard reconstruction 78% RR-interval, motion compensated reconstruction (MCR) significantly improves subjective image quality (SIQ) and diagnostic interpretability across heart rate categories. Clinical relevance By integrating multi-phase data into an optimized single-phase reconstruction with improved SIQ and diagnostic interpretability, MCR may reduce the need for multi-phase assessments when the target phase is non-diagnostic.