Accelerated Segmented Diffusion-Weighted Prostate Imaging for Higher Resolution, Better Geometric Fidelity, and Multi-b Perfusion Quantification
Pelin Aksit Ciris1, Jr-yuan George Chiou2, Andriy Fedorov2, Hualei Shelley Zhang2, Clare Mary Tempany-Afdhal2, Bruno Madore2, and Stephan Ernst Maier2,3

1Department of Biomedical Engineering, Akdeniz University, Antalya, Turkey, 2Department of Radiology, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA, United States, 3Department of Radiology, University of Gothenburg, Gothenburg, Sweden

Synopsis

An accelerated multi-shot diffusion imaging sequence and reconstruction scheme was developed for prostate imaging, allowing improvements in spatial resolution, geometric-fidelity and b-factor coverage to be achieved within a short scan time. Two-fold improvement in spatial resolution and three-fold improvement in geometric fidelity were obtained as compared to single-shot EPI, in twelve prostate cancer patients. In contrast to the standard protocol, which involves separate scans with high (b=1400 and 0 s/mm2) and intermediate (b=500 and 0 s/mm2) diffusion weighting, the proposed accelerated protocol yielded an additional 8 b-factors in half the scan time (5 min 43 s vs. 11 min 48 s).

Purpose

To improve the speed, spatial-resolution and geometric-fidelity of diffusion-weighted imaging in the prostate, and evaluate the possibility of perfusion quantification with multiple b-factors in the same session.

Introduction

Prostate cancer (PCa) is very common and affects approximately one man in every six. DWI is an essential component of PCa diagnosis and staging [1, 2], however, suffers from low resolution, geometric distortions and also low SNR unless signal averaging is employed. Although powerful methods such as multi-coil acceleration can be applied to reduce distortions, this is not practical in the case of prostate imaging with a single, endo-rectal RF coil. A recent accelerated multi-shot acquisition method [3] which is fully compatible with a single coil configuration, exploits the sparsity of diffusion-encoded data in the x-y-kb-kd space, where kb and kd are Fourier-transform duals of b and d, the b-factor and the diffusion direction, respectively. It displaces aliasing artefacts toward underused regions of the kb-kd plane, and recovers non-aliased signals, at potentially no cost in scan time. The approach shifts sampling along the phase-encoding (PE) direction between b-factors and directions, reducing the number of PE steps for each image but not the total number of images. The acceleration scheme allows for relatively good resolution with good geometrical fidelity in a short scan time and also the acquisition of many b-factors, which may enable perfusion quantification and tumour characterization [4-6].

Methods

Twelve patients undergoing PCa staging participated in this IRB-approved study (ages: 62±7 years). Imaging was performed at 3 Tesla (MR750w system, GE Healthcare) using an endo-rectal coil (Medrad). T2-weighted imaging was followed by conventional and accelerated diffusion imaging protocols (Figure 1). For accelerated multi-shot DWI, the segmentation factor equaled the acceleration factor (R=4); thus for each diffusion encoding direction and each b-factor 32 echoes were sampled. Data were reconstructed in the manner introduced in [3], using the magnitude and phase data from a concurrently acquired low-resolution 2D navigator echo (matrix=32x32, TE=128.4ms) for regularization and motion correction, respectively. In order to obtain the highest quality diffusion-weighted images at the diagnostically relevant b-factors 500 and 1400 s/mm2 together with a range of b-factors suitable for multi-exponential analysis, a protocol with variable density b-factor sampling was designed. Moreover, to ensure correct diffusion encoding at very low b-factors, a crusher gradient-free design was employed with a minimum b-factor of 12.5 s/mm2 for adequate alternate signal crushing. Lesions and normal tissue in the peripheral zone and central gland were delineated on conventional (b=500 and 1400 s/mm2) and accelerated direction-averaged DWI. ADC maps were calculated from non-linear least-square fits with monoexponential functions: $$$S/S_0 = exp(-bD)$$$ using multiple b-factors up to b=500 or b=1400 s/mm2 weighted according to the number of averages. Perfusion-free diffusion coefficients (D) and perfusion fractions (f) were calculated in each ROI from non-linear weighted least-square fits with biexponential functions: $$$S/S_0 = (1-f) exp(-bD) + f exp(-bD^*)$$$ [7], using multiple b-factors up to b=500 s/mm2 with D* as a free parameter or fixed at 10 mm2/ms.

Results

Nine patients had lesions; five had a Prostate Imaging Reporting and Data System (PI-RADS) score of 5, four PI-RADS 4, one PI-RADS 2, and two a PIRADS score of 1. Accelerated DWI produced diagnostic quality high-resolution images (Figs. 2 and 3), except for one case with severe motion. Geometric-fidelity improved by a factor of 2.80 ± 1.99 over conventional DWI (Figure 4). Sample bi-exponential fits are shown in Figure 5: a higher perfusion fraction was found in lesions relative to normal tissue.

Discussion

Scan time for the present accelerated acquisition, with 11 b-factors acquired at two-fold resolution and three-times the geometric-fidelity of a single-shot EPI sequence, was below 6 min. In contrast, acquiring 2 single b-factors and b0 took nearly 12 min with single-shot EPI. Geometric-fidelity improved nearly as much as the theoretically-expected 3.33-fold. Perfusion parameters were in good agreement with the limited available literature (f: 3.7-23%, D*: 7-21 μm2/ms, D: 1.2-1.9 μm2/ms in [4, 8-10]) and subsequent DCE acquisitions. These improvements were achieved with a single-channel coil and further improvements in accelerated DWI quality may be possible with the addition of multi-channel coils. Future work includes investigating the effect of regularization strength on calculated ADC values [3] and its effect on residual ghosting such as sometimes seen from the rectal wall.

Conclusion

Accelerated DWI in the prostate with higher spatial-resolution, geometric-fidelity, and multiple b-factors for potential perfusion quantification within the same session, was feasible while maintaining good SNR and reasonable scan times.

Acknowledgements

NIH R01CA160902, R01EB010195, 5R25CA089017-10, P41EB015898 and R01CA149342.

References

1. Tempany, C.M., et al., Multimodal imaging for improved diagnosis and treatment of cancers. Cancer, 2014. doi: 10.1002/cncr.29012.

2. Hegde, J.V., et al., Multiparametric MRI of prostate cancer: an update on state-of-the-art techniques and their performance in detecting and localizing prostate cancer. Journal of Magnetic Resonance Imaging, 2013. 37(5): p. 1035-54.

3. Madore, B., et al., Accelerated multi-shot diffusion imaging. Magn Reson Med, 2014. 72(2): p. 324-36.

4. Dopfert, J., et al., Investigation of prostate cancer using diffusion-weighted intravoxel incoherent motion imaging. Magn Reson Imaging, 2011. 29(8): p. 1053-8.

5. Shinmoto, H., et al., Biexponential apparent diffusion coefficients in prostate cancer. Magn Reson Imaging, 2009. 27(3): p. 355-9.

6. Zhang, Y.D., et al., The histogram analysis of diffusion-weighted intravoxel incoherent motion (IVIM) imaging for differentiating the gleason grade of prostate cancer. Eur Radiol, 2015. 25(4): p. 994-1004.

7. Le Bihan, D., et al., Separation of diffusion and perfusion in intravoxel incoherent motion MR imaging. Radiology, 1988. 168(2): p. 497-505.

8. Pang, Y., et al., Intravoxel incoherent motion MR imaging for prostate cancer: an evaluation of perfusion fraction and diffusion coefficient derived from different b-value combinations. Magn Reson Med, 2013. 69(2): p. 553-62.

9. Mazaheri, Y., et al., Reducing the influence of b-value selection on diffusion-weighted imaging of the prostate: evaluation of a revised monoexponential model within a clinical setting. J Magn Reson Imaging, 2012. 35(3): p. 660-8.

10. Riches, S.F., et al., Diffusion-weighted imaging of the prostate and rectal wall: comparison of biexponential and monoexponential modelled diffusion and associated perfusion coefficients. NMR Biomed, 2009. 22(3): p. 318-25.

Figures

Fig.1: Conventional and accelerated DWI protocols.

Fig.2: Sample conventional and accelerated direction-averaged DWI images with b=500 and 1400 s/mm2 of two patients. The improved spatial resolution with the accelerated scan can clearly be appreciated. To ensure adequate SNR for the accelerated scans, four measurements were obtained at b=500 s/mm2 and eight at b=1400 s/mm2.

Fig.3: Conventional DWI ADC and accelerated DWI diffusion coefficients, D, from single-exponential fits.

Fig.4: (a) Anatomy is severely distorted in conventional DWI, and largely restored using accelerated DWI (c) Distortion was assessed based on apparent prostate length in the AP direction. Geometric-fidelity improved by 280% ± 199, using T2W as the gold-standard, as summarized in Bland-Altman plots.

Fig.5: (a) Sample accelerated DWI images and DCE subtractions from two patients. Bi-exponential fits reveal higher perfusion fractions in both lesions (red), relative to normal tissue (blue). (b) Bi-exponential perfusion-free diffusion coefficients, pseudo-diffusion coefficient and perfusion fractions, for both patients.



Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)
3860