Abstract
Background
Deep learning (DL)-based reconstruction may decrease the scan time of three-dimensional (3D) T1-weighted (T1W) imaging without compromising image quality. However, DL-based super-resolution reconstruction of volumetric interpolated breath-hold examination (DL-VIBE) has not previously been applied to intracranial contrast-enhancing lesions.
Purpose
To investigate the diagnostic performance of DL-VIBE and compare it with conventional 3D T1 magnetization-prepared rapid-acquisition gradient echo (MPRAGE) for assessing intracranial enhancing lesions.
Material and Methods
In this study, 97 patients (35 men, 62 women; mean age = 59.2 ± 15.3 years) who underwent both contrast-enhanced 3D T1W imaging in the same imaging session (1 min 49 s vs. 5 min 32 s) between May and December 2023 were retrospectively included. Two neuroradiologists independently evaluated image quality, gray-white matter differentiation, lesion conspicuity, and artifacts using a 5-point Likert scale. Quantitative metrics included the number and maximum diameter of enhancing lesions, contrast:noise ratio (CNR) of lesion-to-normal parenchyma (CNRlesion/parenchyma), and CNR of white:gray matter (CNRWM/GM).
Results
Although DL-VIBE demonstrated lower overall image quality and gray-white matter differentiation, it showed significantly higher CNRlesion/parenchyma and fewer motion and pulsation artifacts (P <0.001) than conventional MPRAGE. Furthermore, no significant differences were observed in the lesion conspicuity, number, or longest diameter of enhancing lesions between DL-VIBE and conventional MPRAGE (P >0.05).
Conclusion
DL-VIBE showed comparable diagnostic performance for intracranial enhancing lesions with reduced scan time compared to conventional MPRAGE, although it was limited by a lower image quality. Therefore, DL-VIBE represents a promising approach for the technical development of DL-based reconstruction in clinical practice.
Keywords
Get full access to this article
View all access options for this article.