Evaluation of moyamoya disease in CT angiography using ultra-high-resolution computed tomography: Application of deep learning reconstruction

Purpose: The aim of this study was to examine the evaluation of ultra-high-resolution computed tomography angiography (UHR CTA) images in moyamoya disease (MMD) reconstructed with hybrid iterative reconstruction (Hybrid-IR), model-based iterative reconstruction (MBIR), and deep learning reconstruction (DLR).

Methods: This retrospective study with institutional review board approval included patients with clinically suspected MMD who underwent UHR CTA between January 2018 and July 2020. CTA images were reconstructed with three reconstruction methods. Qualitative visualization was evaluated in comparison with digital subtraction angiography. Quantitative evaluation included assessment of edge sharpness, full width at half maximum (FWHM), vessel contrast, and tissue signal-to-noise ratio (SNR_tissue). One-way analysis of variance was used to analyze differences. In addition, reconstruction time were assessed.

Results: Qualitative evaluation of CTA for 33 sides did not differ significantly between reconstruction methods. In quantitative evaluation for 54 patients, edge sharpness for right and left cortical segments of the middle cerebral artery was significantly higher for Hybrid-IR than for other reconstructions. No significant difference was seen between MBIR and DLR. Edge sharpness for STA-MCA bypass was significantly higher for Hybrid-IR than for MBIR, but no significant difference was seen between Hybrid-IR and DLR. FWHM for STA-MCA showed no significant difference between the three reconstruction methods. DLR displayed the highest SNR_tissue. The time required for reconstruction was 40 s for Hybrid-IR, 2580 s for MBIR, and 180 s for DLR.

Conclusion: UHR CTA with DLR adequately visualized vessels in patients with MMD within a clinically feasible reconstruction time.