Clinical Implications of CT-detected Hypoattenuation Thickening on Left Atrial Appendage Occlusion Devices.

Background At follow-up CT after left atrial appendage occlusion (LAAO), hypoattenuation thickening (HAT) on the atrial aspect of the device is a common finding but the clinical implications require further study. Purpose To assess the association of HAT grade at follow-up CT with clinical characteristics and outcomes in patients who underwent LAAO. Materials and Methods This prospective study included consecutive participants with atrial fibrillation and who were at high risk for stroke (CHA2DS2-VASc score ≥4) who underwent LAAO and were administered pacifier or nonpacifier devices at two French medical centers between January 2012 and November 2020. Postprocedure CT images were evaluated by two radiologists in consensus and device-specific interpretation algorithms were applied to classify HAT as low grade (low suspicion of thrombosis) or high grade (high suspicion of thrombosis). The association between HAT grade and clinical characteristics was assessed using multinomial logistic regression, and variables associated with risk of stroke were assessed using a Cox proportional hazard model. Results This study included 412 participants (mean age, 76 years ± 8 [SD]; 284 male participants) who underwent follow-up CT at a mean of 4.2 months ± 1.7 after LAAO. Low-grade and high-grade HAT were depicted in 98 of 412 (23.8%) and 21 of 412 (5.1%) participants, respectively. High-grade HAT was associated with higher odds of antithrombotic drug discontinuation during follow-up (odds ratio, 9.5; 95% CI: 3.1, 29.1; P < .001), whereas low-grade HAT was associated with lower odds of persisting left atrial appendage patency (odds ratio, 0.46; 95% CI: 0.27, 0.79; P = .005). During a median follow-up of 17 months (IQR, 11-41 months), stroke occurred in 24 of 412 (5.8%) participants. High-grade HAT was associated with stroke (hazard ratio, 4.6; 95% CI: 1.5, 14.0; P = .008) and low-grade HAT (P = .62) was not. Conclusion Low-grade HAT was a more common finding at CT performed after LAAO CT (24%) than was high-grade HAT (5%), but it was associated with more favorable outcomes than high-grade HAT, which was associated with higher stroke risk. © RSNA, 2023 Supplemental material is available for this article. See also the editorial by Choe in this issue.

Imaging of the post-operative hallux valgus: what do radiologists need to know?

Hallux valgus surgery concerns many patients and various techniques are performed. The assessment of the first toe deformity correction is mainly visual and imaging is required to analyze the intermetatarsal angle and depict complications. However, it is often difficult for the radiologist to distinguish normal and pathological conditions, especially in case of osteotomies which may show various aspects of bone mineralization and healing. In this review, the most relevant imaging features of the post-operative hallux valgus are summarized.

High-resolution Free-breathing late gadolinium enhancement Cardiovascular magnetic resonance to diagnose myocardial injuries following COVID-19 infection.

PURPOSE: High-resolution free-breathing late gadolinium enhancement (HR-LGE) was shown valuable for the diagnosis of acute coronary syndromes with non-obstructed coronary arteries. The method may be useful to detect COVID-related myocardial injuries but is hampered by prolonged acquisition times. We aimed to introduce an accelerated HR-LGE technique for the diagnosis of COVID-related myocardial injuries. METHOD: An undersampled navigator-gated HR-LGE (acquired resolution of 1.25 mm3) sequence combined with advanced patch-based low-rank reconstruction was developed and validated in a phantom and in 23 patients with structural heart disease (test cohort; 15 men; 55 ± 16 years). Twenty patients with laboratory-confirmed COVID-19 infection associated with troponin rise (COVID cohort; 15 men; 46 ± 24 years) prospectively underwent cardiovascular magnetic resonance (CMR) with the proposed sequence in our center. Image sharpness, quality, signal intensity differences and diagnostic value of free-breathing HR-LGE were compared against conventional breath-held low-resolution LGE (LR-LGE, voxel size 1.8x1.4x6mm). RESULTS: Structures sharpness in the phantom showed no differences with the fully sampled image up to an undersampling factor of x3.8 (P > 0.5). In patients (N = 43), this acceleration allowed for acquisition times of 7min21s ± 1min12s at 1.25 mm3 resolution. Compared with LR-LGE, HR-LGE showed higher image quality (P = 0.03) and comparable signal intensity differences (P > 0.5). In patients with structural heart disease, all LGE-positive segments on LR-LGE were also detected on HR-LGE (80/391) with 21 additional enhanced segments visible only on HR-LGE (101/391, P < 0.001). In 4 patients with COVID-19 history, HR-LGE was definitely positive while LR-LGE was either definitely negative (1 microinfarction and 1 myocarditis) or inconclusive (2 myocarditis). CONCLUSIONS: Undersampled free-breathing isotropic HR-LGE can detect additional areas of late enhancement as compared to conventional breath-held LR-LGE. In patients with history of COVID-19 infection associated with troponin rise, the method allows for detailed characterization of myocardial injuries in acceptable scan times and without the need for repeated breath holds.

Deep Learning Framework for Real-Time Estimation of in-silico Thrombotic Risk Indices in the Left Atrial Appendage.

Patient-specific computational fluid dynamics (CFD) simulations can provide invaluable insight into the interaction of left atrial appendage (LAA) morphology, hemodynamics, and the formation of thrombi in atrial fibrillation (AF) patients. Nonetheless, CFD solvers are notoriously time-consuming and computationally demanding, which has sparked an ever-growing body of literature aiming to develop surrogate models of fluid simulations based on neural networks. The present study aims at developing a deep learning (DL) framework capable of predicting the endothelial cell activation potential (ECAP), an in-silico index linked to the risk of thrombosis, typically derived from CFD simulations, solely from the patient-specific LAA morphology. To this end, a set of popular DL approaches were evaluated, including fully connected networks (FCN), convolutional neural networks (CNN), and geometric deep learning. While the latter directly operated over non-Euclidean domains, the FCN and CNN approaches required previous registration or 2D mapping of the input LAA mesh. First, the superior performance of the graph-based DL model was demonstrated in a dataset consisting of 256 synthetic and real LAA, where CFD simulations with simplified boundary conditions were run. Subsequently, the adaptability of the geometric DL model was further proven in a more realistic dataset of 114 cases, which included the complete patient-specific LA and CFD simulations with more complex boundary conditions. The resulting DL framework successfully predicted the overall distribution of the ECAP in both datasets, based solely on anatomical features, while reducing computational times by orders of magnitude compared to conventional CFD solvers.