Department of Radiologic Technology, Faculty of Associated Medical Sciences, Chiang Mai University, 110 Intavaroros Road, Muang, Chiang Mai, 50200, Thailand.
Department of Radiology, Faculty of Medicine, Chiang Mai University, Intavaroros Road, Muang, Chiang Mai, Thailand.
MAGMA. 2024 Jul;37(3):465-475. doi: 10.1007/s10334-024-01165-8. Epub 2024 May 17.
This study investigated the feasibility of using deep learning-based super-resolution (DL-SR) technique on low-resolution (LR) images to generate high-resolution (HR) MR images with the aim of scan time reduction. The efficacy of DL-SR was also assessed through the application of brain volume measurement (BVM).
In vivo brain images acquired with 3D-T1W from various MRI scanners were utilized. For model training, LR images were generated by downsampling the original 1 mm-2 mm isotropic resolution images. Pairs of LR and HR images were used for training 3D residual dense net (RDN). For model testing, actual scanned 2 mm isotropic resolution 3D-T1W images with one-minute scan time were used. Normalized root-mean-square error (NRMSE), peak signal-to-noise ratio (PSNR), and structural similarity (SSIM) were used for model evaluation. The evaluation also included brain volume measurement, with assessments of subcortical brain regions.
The results showed that DL-SR model improved the quality of LR images compared with cubic interpolation, as indicated by NRMSE (24.22% vs 30.13%), PSNR (26.19 vs 24.65), and SSIM (0.96 vs 0.95). For volumetric assessments, there were no significant differences between DL-SR and actual HR images (p > 0.05, Pearson's correlation > 0.90) at seven subcortical regions.
The combination of LR MRI and DL-SR enables addressing prolonged scan time in 3D MRI scans while providing sufficient image quality without affecting brain volume measurement.
本研究旨在探讨基于深度学习的超分辨率（DL-SR）技术在低分辨率（LR）图像上的可行性，以生成高分辨率（HR）MR 图像，从而减少扫描时间。还通过脑容量测量（BVM）评估了 DL-SR 的效果。
使用来自不同 MRI 扫描仪的 3D-T1W 采集体内脑图像。为了进行模型训练，通过对原始 1mm-2mm 各向同性分辨率图像进行下采样生成 LR 图像。使用 LR 和 HR 图像对来训练 3D 残差密集网络（RDN）。对于模型测试，使用具有一分钟扫描时间的实际扫描的 2mm 各向同性分辨率 3D-T1W 图像。归一化均方根误差（NRMSE）、峰值信噪比（PSNR）和结构相似性（SSIM）用于模型评估。评估还包括脑容量测量，评估皮质下脑区。
结果表明，与立方插值相比，DL-SR 模型提高了 LR 图像的质量，NRMSE（24.22%比 30.13%）、PSNR（26.19 比 24.65）和 SSIM（0.96 比 0.95）。在体积评估方面，DL-SR 和实际 HR 图像之间在七个皮质下区域没有显著差异（p＞0.05，Pearson 相关系数＞0.90）。
LR MRI 和 DL-SR 的结合可以解决 3D MRI 扫描时间过长的问题，同时提供足够的图像质量，而不会影响脑容量测量。
