MRI‐TRUS registration methodology for TRUS‐guided HDR prostate brachytherapy

Abstract Purpose High‐dose‐rate (HDR) prostate brachytherapy is an established technique for whole‐gland treatment. For transrectal ultrasound (TRUS)‐guided HDR prostate brachytherapy, image fusion with a magnetic resonance image (MRI) can be performed to make use of its soft‐tissue contrast. The MIM treatment planning system has recently introduced image registration specifically for HDR prostate brachytherapy and has incorporated a Predictive Fusion workflow, which allows clinicians to attempt to compensate for differences in patient positioning between imaging modalities. In this study, we investigate the accuracy of the MIM algorithms for MRI‐TRUS fusion, including the Predictive Fusion workflow. Materials and Methods A radiation oncologist contoured the prostate gland on both TRUS and MRI. Four registration methodologies to fuse the MRI and the TRUS images were considered: rigid registration (RR), contour‐based (CB) deformable registration, Predictive Fusion followed by RR (pfRR), and Predictive Fusion followed by CB deformable registration (pfCB). Registrations were compared using the mean distance to agreement and the Dice similarity coefficient for the prostate as contoured on TRUS and the registered MRI prostate contour. Results Twenty patients treated with HDR prostate brachytherapy at our center were included in this retrospective evaluation. For the cohort, mean distance to agreement was 2.1 ± 0.8 mm, 0.60 ± 0.08 mm, 2.0 ± 0.5 mm, and 0.59 ± 0.06 mm for RR, CB, pfRR, and pfCB, respectively. Dice similarity coefficients were 0.80 ± 0.05, 0.93 ± 0.02, 0.81 ± 0.03, and 0.93 ± 0.01 for RR, CB, pfRR, and pfCB, respectively. The inclusion of the Predictive Fusion workflow did not significantly improve the quality of the registration. Conclusions The CB deformable registration algorithm in the MIM treatment planning system yielded the best geometric registration indices. MIM offers a commercial platform allowing for easier access and integration into clinical departments with the potential to play an integral role in future focal therapy applications for prostate cancer.

the potential to play an integral role in future focal therapy applications for prostate cancer.

K E Y W O R D S
HDR brachytherapy, MIM, MRI, prostate cancer, registration, TRUS

| INTRODUCTION
High-dose-rate (HDR) prostate brachytherapy is an established treatment technique, in combination with external beam radiotherapy, for intermediate-and high-risk prostate cancer. 1,2 The current approach to prostate cancer radiotherapy involves the irradiation of the entire gland. In recent years, interest has been mounting in treating the prostate using a focal therapy approach. This can involve either escalating the dominant intraprostatic lesions (DILs) of disease to a higher boost dose while maintaining the dose to the entire prostate, treating half of the prostate (termed "hemigland" treatment), or exclusively treating the DIL(s). A summary of select number of focal brachytherapy studies is provided in Table 1. HDR prostate brachytherapy relies heavily on imaging infrastructure and can be delivered using a variety of imaging workflows including integration with computed tomography (CT), magnetic resonance imaging (MRI), and transrectal ultrasound (TRUS). TRUS guidance in HDR brachytherapy has been widely used, 3 due largely to its cost-effectiveness and availability. Unfortunately, however, softtissue resolution on TRUS imaging is poor, creating challenges in resolving intraprostatic features. MRI, in contrast, excels at soft tissue contrast and has been increasingly incorporated into radiotherapy practices to aid with segmentation of both cancerous targets and organs at risk. Historically, the fusion between MRI and TRUS images has only been possible using cognitive registration; however, software-based tools are gradually being introduced into brachytherapy. 4 While several MRI-TRUS fusion tools have been described in literature, both for targeted prostate biopsy and for brachytherapy applications, [5][6][7][8][9][10][11][12][13][14][15] it is important for centers to independently assess registration methodologies for use in their own clinical workflow.
The registration details along with major results of these MRI-TRUS studies are summarized in Table 2, along with these details in relation to the major results from this study. Differences in MRI specifics, including magnetic field strength and use of endorectal coils, and HDR brachytherapy planning strategies can result in changes to the registration accuracy and necessary precision. This study reports the first commercial solution to fuse the MRI and the TRUS images specifically for HDR brachytherapy available in the MIM treatment planning system (MIM Software Inc., Cleveland OH).
In this work, we evaluate two multimodality image registration methodologies within the MIM treatment planning system to fuse the MRI to the TRUS images: rigid registration (RR) and contourbased (CB) deformable image registration. These are assessed in combination with the Predictive Fusion workflow specific to the MIM treatment planning system, which allows the user to identify the expected location of the TRUS probe during brachytherapy on the MRI and reorient the slices of the MRI perpendicular to the probe angle in an effort to improve the fusion by accounting for patient positioning differences.
Additionally, this study provides a preliminary investigation into the application of this workflow in the context of DIL-based foci therapy. Specifically, exploration into intraprostatic landmark-based approach was investigated using patient-specific landmarks.

2.B | Pre-treatment magnetic resonance imaging
All patients underwent a multiparametric MRI scan in advance of HDR prostate brachytherapy. Multiparametric sequences included T1-and T2-weighted scans, diffusion-weighted imaging, and gadolinium contrast-enhanced sequences performed using a 3-T magnet. No endorectal coil was used. The MRI was resampled in the MIM treatment planning system to 1-mm isotropic resolution to match the superior-inferior resolution of the 3D TRUS scan. The prostate was delineated by a radiation oncologist on the resampled T2-weighted image. An example of an MRI scan from one of the study patients is shown in Figure 1.

2.C | Clinical brachytherapy process
The HDR prostate brachytherapy process at the Tom Baker Cancer Centre, Calgary, AB, follows a TRUS-guided, intraoperative-planned approach. The implant is performed with the patient under spinal anesthesia in an unshielded operating room. The patient is  HDR brachytherapy is exclusively delivered as a component of a combined modality regimen at our center at present. All patients undergo one fraction of HDR brachytherapy (prescription dose of 15 Gy) and a 23-fraction course of external beam radiotherapy (prescription dose of 46 Gy). The two treatment components may be delivered in either order. Hormonal therapy may also be offered.

2.D | Image registration and evaluation
For this retrospective study, the pre-brachytherapy MRI and TRUS acquired during brachytherapy were used to investigate and evaluate the quality of MIM-based image registration for twenty patients.
Both MRI and US datasets had the prostate, urethra, and rectal wall contoured by the radiation oncologist. The image registration workflow using MIM is shown in Figure 2. Two registration techniques for the 20 multimodality (MRI and TRUS) imaging datasets were investigated: RR and CB deformable registration. Each was per- The Predictive Fusion workflow within MIM allows the user to position and orient the TRUS probe on the MRI scan. The software then produces a "resliced" MRI that has been reoriented perpendicular to the TRUS probe. This process is shown in Figure 3. No deformation is introduced in this fusion process. After performing the predictive fusion, a RR or CB registration was performed as previously described.
To evaluate the geometric quality of the four variations of registration strategies for each patient, a distance-based metric (mean distance to agreement [MDA]) and a volume-based metric (Dice similarity coefficient) were used, as per AAPM TG-132 recommendations. 16

3.A | Image registration evaluation
The four registrations investigated are compared in Figure 4  When positioning the TRUS probe on the MRI for the Predictive Fusion workflow, the angle selected was an average ± standard deviation 12 ± 5°from the cranial-caudal axis (head-down, as shown in Figure 3). The intrapatient variation was 3 ± 3°among the three independent angle selections.
The MRI prostate volumes were average ± standard deviation 6 ± 17% larger than the TRUS prostate volumes. The deformed MRI contours following CB registration, however, exhibited far greater consistency with the TRUS volumes, with volume differences of average ± standard deviation −1 ± 4% for both CB and pfCB. All  HDR prostate brachytherapy is commonly performed as an intraoperative procedure. While MRI-based workflows are becoming more common, 19 the high cost and availability of the equipment still limit their availability. Ultrasound, conversely, is cost-effective and widely available and requires minimal space and specialized infrastructure. 19 It has been used extensively in prostate brachytherapy procedures yielding practitioner comfort and experience. MRI-TRUS fusion strategies will allow the use of MRI soft-tissue information in a TRUS-guided procedure, improving the ability for centers to implement targeted strategies while maintaining their TRUS-guided procedure. In this study, MRIs were obtained with wide variability in timing prior to brachytherapy (see Table 3); these images were required for patients' standard clinical care, not specifically for use in this study. The ability of the image fusion algorithm to perform with high consistency on these images is further indication of the strength of the MIM CB deformable image registration algorithm.
Further, in this study, all contouring was done retrospectively on for example) and fractionations remain under investigation.
We elected to use the prostate contours in this study for registration as these are well visualized on both MRI and TRUS imaging modalities. Implementation in our clinical process would thus involve contouring of the prostate intraoperatively. Exceptional agreement of the prostate contour registration metrics, MDA and DICE (Figure 4), lends confidence to this process; as uncertainties were F I G . 1. Sample axial and sagittal TRUS (a) and MRI (b) images from this study. The prostate is contoured in purple and cyan and the urethra is contoured in yellow and red on the TRUS and MRI, respectively.
identified in the base and apex regions, however, critical dosimetric assessment at those locations (particularly in the case of disease located in those locations) would be warranted in an intraoperative setting.
Rotation of the coordinate system when comparing an MRI, which is obtained with the legs down to allow the patient to fit inside the imager bore, and the TRUS, which is obtained in the lithotomy position, has been previously reported in the literature, 11,15 motivating solutions such as the Predictive Fusion option offered by MIM. It has, however, also been reported that including rotation in prostate RRs may degrade registration quality. 9 In this study, the CB deformable registration was performed after three separate manual  This study also provides a preliminary investigation into the performance of this MIM-based workflow for applications that may have implications for DIL-based foci HDR therapy, as well as recommendations to improve the accuracy of implementation and thus reducing the required safety margin to ensure treatment accuracy in foci therapy.
The intraprostatic landmark registration approach showed promise from the preliminary investigation on five of the 20 retrospective patients. All generated an MDA of less than 2 mm when

| CONCLUSIONS
In this study, we have reported the first commercial brachytherapy solution for TRUS-MRI fusion in the MIM treatment planning system. CB deformable registration yielded superior geometric results compared to RR when considering the external prostate contour.
Utilizing the predictive fusion tool yielded similar results for the geometric indices, MDA and DICE, for both rigid and CB DIR in absence of the tool and was not seen as mandatory for the MRI-TRUS fusion workflow. This MRI-TRUS based workflow looks promising for application in HDR foci-therapy for prostate brachytherapy.

ACKNOWLEDG EMENT
The authors would like to acknowledge MIM Software Inc. for the donation of a temporary research license for Predictive Fusion and their assistance developing workflows.

CONFLI CTS OF INTEREST
The authors have no relevant conflicts of interest to disclose. design study, data analysis, review and edit manuscript.

D A T A A V A I L A B I L I T Y S T A T E M E N T
The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.