深度学习重建技术在优化前列腺磁共振T2加权成像扫描时间和图像质量中的应用价值

Value of deep learning reconstruction in optimizing prostate MR T2-weighted imaging scanning time and imaging quality

目的 探讨深度学习重建(deep learning reconstruction, DLR)技术在提高前列腺MRI T2加权成像(T2 weighted imaging, T2WI)图像质量及缩短扫描时间中的应用价值。材料与方法 本研究前瞻连续纳入未经治疗的可疑前列腺病变的受试者,分别行前列腺MRI常规快速自旋回波(fast-spin echo, FSE)-T2WI和DLR快速FSE-T2WI扫描,并保存未应用DLR的原始快速FSE-T2WI。由2名研究者分别对三组T2WI(常规T2WI、快速T2WI和DLR快速T2WI)的整体图像质量和图像伪影进行图像质量主观评价(5分标准)。由1名研究者测量前列腺正常外周带、正常移行带和病变的信噪比(signal-to-noise ratio, SNR)以及与髂腰肌的对比噪声比(contrast-to-noise ratio, CNR)。对正态分布和非正态分布的数据分别进行单因素方差分析和Kruskal-Wallis检验,比较分析三组T2WI图像的主观评分和客观指标的差异。采用组内相关系数(intra-class correlation coefficient, ICC)评估研究者之间主观评分和病灶前列腺影像报告和数据系统2.1版(Prostate Imaging-Reporting and Data System version 2.1, PI-RADS v2.1)评分的一致性。结果 本研究共纳入35名受试者(38个前列腺病灶)。DLR快速FSE-T2WI较常规FSE-T2WI扫描时间缩短了32.1%。两位研究者的评分结果均显示,常规FSE-T2WI、快速FSE-T2WI和DLR快速FSE-T2WI的整体图像质量评分、前列腺包膜显示清晰度和前列腺病变显示清晰度均存在显著差异(P<0.05);但在伪影评分上差异无统计学意义(P>0.05)。三组FSE-T2WI图像的前列腺外周带、移行带和病灶的SNR、CNR间差异具有统计学意义(P<0.05)。应用三组T2WI图像进行前列腺病变的PI-RADS v2.1评分具有很好的一致性。结论 DLR可以显著改善快速采集MRI序列的图像质量,有利于促进前列腺快速MRI序列的临床应用。

Objective:To explore the application of deep learning reconstruction(DLR)in improving prostate MRI T2 weighted imaging(T2WI)quality and shortening scanning time.Materials and Methods:Patients who were suspected with a prostate lesion clinically were prospectively enrolled in this study.Conventional MRI fast-spin echo(FSE)-T2WI sequence and DLR fast FSE-T2WI were performed,and the original fast FSE-T2WI without DLR was preserved.The overall image quality,image artifacts,prostate capsule,prostate lesion detection and the lesion's Prostate Imaging-Reporting and Data System version 2.1(PI-RADS v2.1)scoring of three T2WI(conventional T2WI,fast T2WI,and DLR fast T2WI)were assessed subjectively by two radiologists independently.The signal-to-noise ratio(SNR)of the lesion and the contrast-to-noise ratio(CNR)were measured by one radiologist.One-way ANOVA and Kruskal-Wallis test were performed on normally and non-normally distributed data,respectively,to compare and analyze the differences in subjective scores and objective indices of three T2WI.The intra-class correlation coefficient(ICC)was used to compare the interreader agreement of subjective scores and PI-RADS v2.1 scoring between two radiologists.Results:Finally,a total of 35 patients(38 prostate lesions)were enrolled in this study.DLR fast T2WI reduced 32.1%scanning time than conventional T2WI.Two radiologists'assessment demonstrated that there were significant differences among conventional,fast and DLR FSE-T2WI in overall image quality,prostate capsule demonstration and prostate lesion detection(P<0.05).There were significant differences in the overall image quality,prostate capsule demonstration and prostate lesion detection among the three T2WI(P<0.05).The SNR and CNR of prostate peripheral zone,transition zone and prostate lesion of the three T2WI images were significantly different(P<0.05).DLR fast T2WI has the best overall image quality with the least artifacts and short scan time.Conclusions:DLR can significantly improve the image quality of prostate FSE-T2WI with a shorter scanning time.