Table 1_Deep learning-based reconstruction improves MRI image quality and diagnostic performance for carotid atherosclerotic plaques.docx

ObjectiveTo evaluate the image quality and diagnostic performance of fast magnetic resonance imaging with deep learning-based reconstruction (MRIDLR) for carotid atherosclerotic plaques. We compared MRIDLR to conventional MRI (MRIC) and fast MRI without deep learning-based reconstruction (MRIFast).
MethodsSixty-nine patients were recruited between May 2022 and April 2024. Imaging was performed using a 3.0 T MRI system (SIGNA™ Architect, GE Healthcare) with a 19-channel head–neck coil. Image quality was assessed objectively [measuring signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR)], subjectively (evaluating overall image quality, noise, contrast, artifacts, sharpness and diagnostic performance). All statistical analyses were performed using R software version 4.4.2, with agreement among results assessed using Cohen’s κ.
ResultsThe average scanning time was reduced by 12.4 min (approximately 63.2%) for MRIDLR and MRIFast compared to MRIC. The SNR of T1 and T2 sequences for MRIDLR and MRIC showed significantly higher than MRIFast (p < 0.05), and the SNR of T1 and T2 sequences for MRIDLR was significantly higher than MRIC. The SNR of PD sequences for MRIDLR and MRIC showed significantly higher than MRIFast (p < 0.05), and there was not significantly different between MRIDLR and MRIC (p > 0.05). The CNR of T1, T2, and PD sequences for MRIDLR and MRIC showed significantly higher than MRIFast (p < 0.05), and there was not significantly different between MRIDLR and MRIC (p > 0.05). Subjective image quality was significantly higher for MRIDLR than MRIC and MRIFast (p < 0.05), and MRIC was higher than MRIFast (p < 0.05). Comparing with MRIC and MRIFast, MRIDLR was more effective in diagnosing hemorrhage, lipid nuclei, calcification, fiber cap disruption, and vulnerable plaques, and showed greater concordance with surgical pathology findings.
ConclusionMRIDLR can offer superior image quality and diagnostic performance for carotid atherosclerotic plaques while reducing scanning time. So it enhances diagnostic accuracy and lessens the examination burden on patients which will contribute to the diagnosis and treatment of stroke.