Validation of a deformable MRI to CT registration algorithm employing same day planning MRI for surrogate analysis

Abstract Purpose Validating deformable multimodality image registrations is challenging due to intrinsic differences in signal characteristics and their spatial intensity distributions. Evaluating multimodality registrations using these spatial intensity distributions is also complicated by the fact that these metrics are often employed in the registration optimization process. This work evaluates rigid and deformable image registrations of the prostate in between diagnostic‐MRI and radiation treatment planning‐CT by utilizing a planning‐MRI after fiducial marker placement as a surrogate. The surrogate allows for the direct quantitative analysis that can be difficult in the multimodality domain. Methods For thirteen prostate patients, T2 images were acquired at two different time points, the first several weeks prior to planning (diagnostic‐MRI) and the second on the same day as the planning‐CT (planning‐MRI). The diagnostic‐MRI was deformed to the planning‐CT utilizing a commercially available algorithm which synthesizes a deformable image registration (DIR) algorithm from local rigid registrations. The planning‐MRI provided an independent surrogate for the planning‐CT for assessing registration accuracy using image similarity metrics, including Pearson correlation and normalized mutual information (NMI). A local analysis was performed by looking only within the prostate, proximal seminal vesicles, penile bulb, and combined areas. Results The planning‐MRI provided an excellent surrogate for the planning‐CT with residual error in fiducial alignment between the two datasets being submillimeter, 0.78 mm. DIR was superior to the rigid registration in 11 of 13 cases demonstrating a 27.37% improvement in NMI (P < 0.009) within a regional area surrounding the prostate and associated critical organs. Pearson correlations showed similar results, demonstrating a 13.02% improvement (P < 0.013). Conclusion By utilizing the planning‐MRI as a surrogate for the planning‐CT, an independent evaluation of registration accuracy is possible. This population provides an ideal testing ground for MRI to CT DIR by obviating the need for multimodality comparisons which are inherently more challenging.

techniques in an attempt to overcome the limitations of rigid registrations. 2,3 DIR in the prostate is inherently challenging due to a variety of factors including significant variation in anatomy due to variability in rectal and bladder filling, differences in patient positioning, and incomplete knowledge and modeling of how these tissues deform over time and motion. Specifically regarding MRI to CT deformation in prostate, there are significant differences in the properties of MRI and CT imaging datasets. 2,4 For DIRs, several strategies have been developed to characterize and quantify DIR algorithms. 3,5,6 This is an area of active development but several tools exist to characterize and validate DIRs, which include physical phantoms, 7,8 digital phantoms, 9,10 and anatomical landmarks for validation. 11,12 In the multimodality DIR setting, fewer validation strategies exist and creating them is even more challenging. Presented here is a novel method for evaluation of multimodality registrations of the prostate between diagnostic-MRI and radiation treatment planning-CT by utilizing a planning-MRI after fiducial marker placement as a surrogate. By using the surrogate, direct quantitative analysis utilizing spatial intensity-based metrics can be employed which otherwise would be difficult to implement in multimodality settings.

2.B | Imaging
All patients underwent a diagnostic mpMRI study approximately 1 month prior to radiation planning-CT. The mpMRI includes T2-weighted, Dynamic Contrast Enhanced (DCE) and Apparent Diffusion Coefficient datasets. All mpMRI sequences were acquired with size and spacing suitable for registration with the planning-CT.
MRI exams were carried out on a Discovery-MR750 3T-MRI (General Electric; Chicago, Illinois). For the purposes of this study, only the T2-weighted sequence was used. The axial T2w-MRI has a resolution 1.25 9 1.25 9 2.5 mm 3 , Field of View: 320 9 320 mm 2 ; slice thickness; 2.5 mm; 72 slices; repetition time 5500 ms and echo time 100 ms.
The diagnostic-MRI was used to delineate the dominant lesion(s) in the prostate and provide targets for MRI-Ultrasound fused prostate biopsy 13 and later to plan the RT tumor boost. 14 Table 1. The planning-CT was acquired from the diaphragm to mid femur at a slice thickness of 2.0 mm.
Following the planning-CT, a planning-MRI was collected, typically within an hour to maximize similarity between these datasets.
The planning-MRI exam consists of a T2-weighted study, a T2* fast gradient-echo study for visualizing the gold fiducials and several other imaging studies that are not utilized in this analysis. The PADGETT ET AL.
| 259 T2-weighted study is collected with identical parameters as the diagnostic-MRI study. The T2*-weighted study, MERGE TM , is acquired with the same size and spacing to match the T2-weighted datasets.

2.D | Registration evaluation and statistics
The T2-weighted acquisition from the planning-MRI study was employed as the surrogate for the planning-CT for analysis of both T A B L E 1 Summary of patient information (left) and discrepancy (mm) of fiducial alignment between planning-MRI and planning-CT (right).

Subject information summary
Plan-MRI to plan-CT rigid fiducial accuracy (mm)  combined), and the expanded structure (the combined structure expanded by 5 mm). The combined and expanded structures were included for determination of regional registration quality.

| RESULTS
Alignment of planning-MRI to planning-CT was confirmed to be sub-  Figure 3 shows an example of the conformity of the anatomy between the planning-CT and the deformed MRI. other body sites may also benefit from this approach. Specifically, sites where MRI is often incorporated into the treatment planning process and where rigid registrations are frequently suboptimal; abdomen, head and neck, brain pre/postsurgery, etc.
The DIR validation technique described in this manuscript has several unique attributes. The data employed were collected from protocol patients and do not use simulated images or images of artificial materials. This has the benefit of testing the DIR algorithm using images collected on humans using the equipment present in the clinic, thus reflecting the clinic workflow. Another advantage is that none of the datasets are simulated or altered prior to DIR, thus eliminating any potential issues of utilizing artificial materials or simulated data. One challenge with this validation technique is that it lacks a known DVF to compare the resulting deformable registration to and instead relies on correlations between datasets. While a known DVF is a robust solution, the correlation metrics implemented here share the ability to evaluate registration accuracy across any region that is defined by the user, albeit not pixel by pixel. The emergence of multimodality image deformable registrations holds great promise and will facilitate a more seamless integration of MRI and other imaging modalities into the RT planning process among other applications outside of radiation oncology.
In order for multimodality DIRs to be widely adopted, robust validation of these algorithms is necessary. This unique method of validation of multimodality registration utilizing a planning-MRI as a surrogate complements existing validation methods. Piper also has an ownership interest in MIM Software Inc.