Presenter(s)
J. S. Kim1,2, C. C. Conlin3, M. A. Anderson4, T. Barrett5, S. Do1, M. E. Hahn3, M. Harisinghani4, D. Margolis6, R. Marks3, P. Murphy3, N. Nakrour4, M. Ohliger7, R. Rakow-Penner3, M. Rojo Domingo1,8, Y. Song1,9, N. Wehrli6, S. A. Woolen7, T. M. Seibert1,8, and A. M. Dale10; 1Department of Radiation Medicine and Applied Sciences, University of California San Diego, La Jolla, CA, 2Loma Linda University School of Medicine, Loma Linda, CA, 3Department of Radiology, University of California San Diego, La Jolla, CA, 4Department of Radiology, Massachusetts General Hospital, Boston, MA, 5Department of Radiology, Addenbrooke's Hospital, Cambridge University Hospitals NHS Foundation Trust, Cambridge, United Kingdom, 6Department of Radiology, Weill Cornell Medicine, New York, NY, 7Department of Radiology and Biomedical Imaging, University of California San Francisco, San Francisco, CA, 8Department of Bioengineering, University of California San Diego, La Jolla, CA, 9Department of Electrical and Computer Engineering, University of California San Diego, La Jolla, CA, 10Department of Multimodal Imaging & Genetics, J. Craig Venter Institute, San Diego, CA
Purpose/Objective(s): Diffusion-weighted MRI (DWI) is useful for identifying prostate cancer (PC) for focal radiotherapy boost, yet its utility for radiotherapy targeting is limited as standard DWI using echo-planar imaging (EPI) is prone to anatomical distortion artifacts. Previous distortion correction techniques, like the Reversed Polarity Gradient (RPG) method and FSL topup, aim to estimate distortion from two EPI volumes acquired with opposite phase-encoding polarity. However, they also attempt to correct for signal intensity changes from voxel overlap/stretching, which can result in poor estimation of the underlying tissue deformation. We propose an approach that eliminates these confounding signal-intensity changes prior to estimating distortion. We hypothesize that this method will yield improved distortion correction compared to previous techniques.
Materials/Methods: Our proposed method, multi-b RPG (mRPG), relies on multiple pairs of opposite-polarity DWI volumes with different b-values. Normalizing signal intensity across multiple b-values eliminates signal-intensity changes from voxel stretch/overlap. Tissue deformation is then estimated by registering the normalized volumes of opposite polarity. In this retrospective, IRB-approved study, patients with known or suspected PC were included if scanned using two multi-b-value DWI series with opposite phase-encoding polarity. Degree of anatomic distortion was quantified by the correlation between opposite-polarity DWI images; less distortion yields higher correlation. Tissue displacement was also measured in mm along the posterior prostate border and for voxels within target lesions. The distribution of correlation coefficients was examined before and after correction by RPG, topup, and mRPG. Correlation between repeated scans of the same polarity was computed to establish a benchmark value. Median and 10th percentile for each distribution were reported with 1000-sample bootstrap 95% confidence intervals.
Results: 1180 patient scans were available for analysis. Maximum tissue displacement ranged 0.5 – 13.8 mm for the posterior prostate border and 0.5 – 7.5 mm for target lesions. Displacement of the posterior prostate border = 4 mm was observed in 29% of patients. mRPG yielded significantly improved correlation between opposite-polarity DWI compared to other methods (Table).
Conclusion: mRPG demonstrated superior correction of anatomical distortion compared to previous methods. Anatomically accurate DWI could facilitate its use for focal radiotherapy tumor boost.
| Opposite polarity | Same polarity benchmark | ||||
| Uncorrected | RPG | Topup | mRPG | ||
| 10th percentile | .71 [.70 .72] | .74 [.73 .74] | .68 [.68 .70] | .83 [.82 .84] | .84 [.82 .85] |
| Median | .83 [.82 .83] | .84 [.84 .84] | .81 [.81 .82] | 0.90 [.89 .90] | .89 [.88 .89] |