Jyh-Miin Lin1, Andrew J. Patterson2, Chung-Wei Lee3, Ya-Fang Chen3, Tilak Das4, Daniel Scoffings5, Hsiao-Wen Chung6, Jonathan H. Gillard1, and Martin J. Graves2
1Department of Radiology, University of Cambridge, Cambridge, United Kingdom, 2MRIS unit, Cambridge University Hospitals NHS Foundation Trust, Cambridge, United Kingdom, 3Department of Radiology, National Taiwan University College of Medicine, Taipei, Taiwan, 4Addenbrooke’s Hospital, Cambridge, United Kingdom, 5Department of Radiology, Cambridge University Hospitals NHS Foundation Trust, Cambridge, United Kingdom, 6Department of Electrical Engineering, National Taiwan University, Taipei, Taiwan
Synopsis
PROPELLER MRI can reduce motion artifacts. However, the colored noise of PROPELLER could degrade the image quality. Although the iterative Pseudo-Inverse with COnstraints (PICO) has been proposed to improve the image quality metrics, further clinical validation is needed. In this study, two neuroradiologists compared the image quality of PICO with the standard density compensation. Results show that PICO significantly improves the identification of two anatomical structures and the clinical utility. PURPOSE
PROPELLER MRI can reduce the physiological motion artifacts, but the standard density compensation (DC) reconstruction method for PROPELLER leads to suboptimal SNR performance. Such SNR degradation is caused by the “colored noise”.
1 To improve the image quality, we have recently implemented a reconstruction method, namely the Pseudo-Inverse with
COnstraints (PICO), to mitigate the effects of color noise.
2 Although PICO can improve the objective image quality,
3 clinical validation is required to subjectively evaluate the quality of PICO. The purpose of this study is to compare the image quality of PICO with the standard DC, providing clinical evaluations carried out by experienced neuroradiologists.
METHODS
The PICO reconstruction scheme allows for the inclusion of total variation constraint, which can reduce the regional spatially varying fluctuations. PICO solves the following constrained optimization problem:
$$x=argmin(TV(x)) \ such \ that \ \parallel y - Ax \parallel _2 ^2 < \sigma ^2$$
where $$$x$$$ is the resulting image; $$$TV$$$ is total variation, $$$y$$$ is the data, $$$A$$$ is the encoding matrix that represent the PROPELLER trajectory, $$$\sigma ^2$$$ is the maximum tolerated error.
Informed consents for the ethically approved study were obtained from eight normal volunteers (three males and five females, with an average age of 26). T2-weighted fast spin-echo PROPELLER brain data were acquired with a matrix size of 512512 on a 1.5T system (MR450, GE Healthcare, Waukesha, WI, USA), using an 8-channel head array. The imaging parameters were: TR/TE = 2500/85.5ms, echo-spacing = 6.8 ms, ETL = 24, receiver bandwidth = ±125 kHz. FOV = 26cm, 12 slices, and 20 blades, corresponding to the total scanning times of 55s. Raw data were reconstructed by PICO and the standard DC for comparison.In vivo images were rated by two experienced neuroradiologists who were blind to the acquisition protocol and reconstruction methods. The evaluation was carried out on an Agfa IMPAX 6.5.2.114 workstation equipped with a reading monitor of BARCO E-3620 MA which is a 3 MegaPixel flat panel display, which is calibrated twice per year. Images reconstructed from two reconstruction methods were visually compared The two raters independently graded these images. The “clinical utility” was graded on a five-point scale (from 1 to 5, with 5 being the highest quality) according to the classification system in a literature.4 The “identification” of three anatomical structures (thalamus, lentiform nucleus, and head of caudate) and the image quality degradation by artifacts were graded on a three-point scale (from 1 to 3, corresponding to poor, fair, and good, respectively).
Statistic analysis was performed using MATLAB (Mathworks, Natick, MA, USA). Inter-rater reliability was assessed using Spearman’s rho. The nonparametric Wilcoxon’s signed rank test was employed to compare PICO reconstruction and DC. A threshold of p-value < 0.05 was defined as statistically significant.
RESULTS
T2 weighted brain images reconstructed by PICO revealed the higher SNR than the standard DC, with small brain structures visible (Fig. 1). The correlation between two raters were moderate (Spearman rho r
s = 0.55). The scores for PICO were significantly higher than DC for the identification of thalamus (p = 0.03) and lentiform nucleus (p = 0.03). However, the score for PICO was only marginally higher than DC in identifying the head of caudate. Regarding the clinical utility, PICO was significantly higher than DC (p < 0.05). While SNR of PICO was higher than DC, both reconstruction methods are immune from the streak artifacts and pulsation artifacts (Fig. 2).
DISCUSSION
PROPELLER images reconstructed by PICO showed superior image quality over the standard DC. The higher SNR of PICO reconstruction is consistent with the better identification of the anatomical structures and the higher clinical utilities. As noise has a detrimental effect on image quality, the denoising property of PICO for PROPELLER MRI could have broad clinical implications.
CONCLUSION
PICO reconstruction increased SNR, thereby improving the identification of anatomical structures and clinical utilities.
Acknowledgements
We thank for the assistance of radiographers at MRIS, Cambridge University Hospitals NHS foundation. Financial supports from Addenbrooke’s Charitable Trust and the NIHR Comprehensive Biomedical Research Centre, in partnership with University of Cambridge are acknowledged. J.M.L received funding from Cambridge Commonwealth, European & International Trust and Ministry of Education, Republic of
China (Taiwan).References
1. Wang F-N, Huang T-Y, Lin F-H, et al. PROPELLER EPI: An MRI technique suitable for diffusion tensor imaging at high field strength with reduced geometric distortions. Magn Reson Med 2005;54(5):1232-1240.
2. Lin J-M, Patterson A, Chang H-C, et al. Whitening of colored noise in PROPELLER using iterative regularized PICO reconstruction. In Proceedings of the 23rd Annual Meeting of ISMRM. Toronto, Canada; 2015. p 3738.
3. Lin J-M, Patterson AJ, Chang H-C, et al. An iterative reduced field-of-view reconstruction for periodically rotated overlapping parallel lines with enhanced reconstruction (PROPELLER) MRI. Med Phys 2015;42(10):5757-5767.
4. Dietrich TJ, Ulbrich EJ, Zanetti M, et al. PROPELLER technique to improve image quality of MRI of the shoulder. AJR Am J Roentgenol 2011;197(6):W1093-1100.