Program Information
A Complete Development of MRI-Only Based Automated Detection Algorithm of Low-Dose Rate Brachytherapy Seeds for Post Implant Dosimetry Applications
R Nosrati1,2*, H Safigholi2 , A Shaaer1 , G Morton2 , A Pejovic-Milic1 , W Song1,2 , (1) Ryerson University, Toronto, ON, (2) Sunnybrook Health Sciences Centre, Toronto, ON
Presentations
SU-K-FS4-3 (Sunday, July 30, 2017) 4:00 PM - 6:00 PM Room: Four Seasons 4
Purpose: To develop a competitively efficient and automated MRI-only based workflow for detecting locations and orientations of LDR brachytherapy seeds for post implant dosimetry applications for prostate cancer (and other sites) that can replace the current CT based approach.
Methods: An agar-based prostate mimicking phantom containing: 20 stranded I-125 dummy seeds in five strands with 0.5mm spacing (IsoAid AdvantageTM) and three calcifications (from sheep bone) with different lengths (8.2-2mm), was constructed. Seed strands were placed at different orientations with respect to main magnetic field. The scan was performed on a 3T Philips Achieva MRI scanner with an 8-channel head array coil using a 3D multi-echo GRE sequence with following parameters: TE1=2ms; number-of-echoes=4; echo spacing=1.5ms; TR=30ms; flip angle=15ᵒ; resolution=0.7mm isotropic. Quantitative Susceptibility Mapping (QSM) was performed using an optimized exponential formulation and MEDI. Number of seeds was calculated by silhouette criterion and seed centroids were determined using K-medoid clustering. Seed orientations were estimated by Eigenvector analysis. The phantom was also scanned using CT and seed locations were identified using VariSeed software. The dose distributions were calculated using Monte Carlo simulation (MCNP-6) and differences between total dose distributions were investigated.
Results: The proposed QSM-based algorithm successfully generated distinctive positive contrast for the seeds which was significantly different from calcifications. No significant difference was found between centroids calculated by VariSeed and those calculated by proposed methods. The estimated orientations were in excellent agreement with actual seed orientations (R>0.95). The difference in the calculated dose distributions showed that the seed orientations (provided only by the proposed MR-based methods) results in a few voxels (<4 voxels for each seed) with severe over- or under-estimated doses.
Conclusion: The proposed workflow has great potential to replace the conventional CT-based approach. The performance of the proposed methods needs to be assessed on patients to facilitate clinical utilization.
Funding Support, Disclosures, and Conflict of Interest: NSERC-CGS-D3
Contact Email: