基于参考图压缩传感方法的前列腺肿瘤定量DCE-MRI

摘要

Objective To investigate the accuracy of quantitative DCE-MR (qDCE-MR) images of prostate tumor reconstructed by usingreference image compressed sensing (RICS) method, as well as the reliability of quantitative pharmacoki-netic parameters of prostate tumor estimated from these images. Methods qDCE-MR images of prostate tumor reconstructed by RICS and the quantitative pharmacokinetic parameters estimated from these images were compared to those reconstructed by conventional fully sampling method. Several quantitative criteria were implemented to evaluate the accuracy and reliability of RICS. Results The mean squared error (MSE) between qDCE-MR images reconstructed by RICS method and the conventional fully sampling method were at a level of 10-5 , meaning that they were very close to each other. The difference between them increased as the acceleration coefficient increasing. Even at an acceleration coefficient of 6, the correlation between quantitative pharmacokinetic parameters estimated from RICS and conventional fully sampling method was still high (Pearson correlation coefficient of Ktrans and vc were r=0. 9260, P<0. 01 and r = 0. 9948, P<0. 01, respectively). Conclusion Reference image compressed sensing based qDCE-MRI of prostate tumor can provide accurate image reconstruction results. Through the accuracy decreased as the acceleration factor increasing, reliable quantitative pharmacokinetic parameters can be estimated at an acceleration factor of 6.%目的 探讨基于参考图压缩传感方法(RICS)的前列腺肿瘤定量DCE-MRI (qDCE-MRI)技术中图像重建的精确性和定量化药物代谢动力学参数估计的可靠性.方法 比较RICS方法和全采样方法的qDCE-MR图像重建结果,以及由此估计的定量化药物代谢动力学参数.通过设置定量化指标,评估基于RICS的前列腺肿瘤qDCE-MRI的精确性和可靠性.结果 RICS法和传统全采样法的qDCE-MR图像重建结果之间的平均平方误差(MSE)在10-5量级,两种重建结果非常接近,但随加速倍数增加,两者差距增大.加速倍数为6时,通过RICS法和全采样法估计出的药物代谢动力学参数仍有较高相关性(两者之间的Ktrans和vc的Pearson相关系数分别为r=0.9260,P＜0.01和r=0.9948,P＜0.01).结论 基于RICS的前列腺肿瘤qDCE-MRI能够给出精确的图像重建结果,重建精确性随加速倍数增大而减小；加速倍数为6时,仍能估计出可靠的定量化药物代谢动力学参数.