Haji-Valizadeh Hassan, Shen Daming, Avery Ryan J, Serhal Ali M, Schiffers Florian A, Katsaggelos Aggelos K, Cossairt Oliver S, Kim Daniel
Departments of Biomedical Engineering (H.H.V., D.S., D.K.), Computer Science (F.A.S.), and Electrical and Computer Engineering (A.K.K., O.S.C.), McCormick School of Engineering, Northwestern University, Evanston, Ill; and Department of Radiology, Feinberg School of Medicine, Northwestern University, 737 N Michigan Ave, Suite 1600, Chicago, IL 60611 (H.H.V., D.S., R.J.A., A.M.S., D.K.).
Radiol Cardiothorac Imaging. 2020 Jun 25;2(3):e190205. doi: 10.1148/ryct.2020190205.
To implement an integrated reconstruction pipeline including a graphics processing unit (GPU)-based convolutional neural network (CNN) architecture and test whether it reconstructs four-dimensional non-Cartesian, non-contrast material-enhanced MR angiographic k-space data faster than a central processing unit (CPU)-based compressed sensing (CS) reconstruction pipeline, without significant losses in data fidelity, summed visual score (SVS), or arterial vessel-diameter measurements.
Raw k-space data of 24 patients (18 men and six women; mean age, 56.8 years ± 11.8 [standard deviation]) suspected of having thoracic aortic disease were used to evaluate the proposed reconstruction pipeline derived from an open-source three-dimensional CNN. For training, 4800 zero-filled images and the corresponding CS-reconstructed images from 10 patients were used as input-output pairs. For testing, 6720 zero-filled images from 14 different patients were used as inputs to a trained CNN. Metrics for evaluating the agreement between the CNN and CS images included reconstruction times, structural similarity index (SSIM) and normalized root-mean-square error (NRMSE), SVS (3 = nondiagnostic, 9 = clinically acceptable, 15 = excellent), and vessel diameters.
The mean reconstruction time was 65 times and 69 times shorter for the CPU-based and GPU-based CNN pipelines (216.6 seconds ± 40.5 and 204.9 seconds ± 40.5), respectively, than for CS (14 152.3 seconds ± 1708.6) (P < .001). Compared with CS as practical ground truth, CNNs produced high data fidelity (SSIM = 0.94 ± 0.02, NRMSE = 2.8% ± 0.4) and not significantly different (P = .25) SVS and aortic diameters, except at one out of seven locations, where the percentage difference was only 3% (ie, clinically irrelevant).
The proposed integrated reconstruction pipeline including a CNN architecture is capable of rapidly reconstructing time-resolved volumetric cardiovascular MRI k-space data, without a significant loss in data quality, thereby supporting clinical translation of said non-contrast-enhanced MR angiograms. © RSNA, 2020.
实施一种集成重建流程，该流程包括基于图形处理单元（GPU）的卷积神经网络（CNN）架构，并测试其重建四维非笛卡尔、非对比剂增强磁共振血管造影k空间数据的速度是否比基于中央处理器（CPU）的压缩感知（CS）重建流程更快，且在数据保真度、综合视觉评分（SVS）或动脉血管直径测量方面无显著损失。
使用24例疑似胸主动脉疾病患者（18例男性和6例女性；平均年龄56.8岁±11.8[标准差]）的原始k空间数据，评估从开源三维CNN派生的拟议重建流程。在训练中，将来自10例患者的4800幅零填充图像和相应的CS重建图像用作输入-输出对。在测试中，将来自14例不同患者的6720幅零填充图像用作训练后CNN的输入。评估CNN与CS图像之间一致性的指标包括重建时间、结构相似性指数（SSIM）和归一化均方根误差（NRMSE）、SVS（3分=无法诊断，9分=临床可接受，15分=优秀）以及血管直径。
基于CPU的CNN流程和基于GPU的CNN流程的平均重建时间分别比CS流程（14152.3秒±1708.6）短65倍和69倍（分别为216.6秒±40.5和204.9秒±40.5）（P<.001）。与作为实际参考标准的CS相比，CNN产生了高数据保真度（SSIM=0.94±0.02，NRMSE=2.8%±0.4），且SVS和主动脉直径无显著差异（P=.25），除了七个位置中的一个位置，其百分比差异仅为3%（即临床无关）。
所提出的包括CNN架构的集成重建流程能够快速重建时间分辨的容积心血管MRI k空间数据，而不会显著损失数据质量，从而支持所述非对比增强磁共振血管造影的临床转化。©RSNA，2020。
