2010 Summer Project Week MR to CT Registration for Prostate Brachytherapy Dose Calculation
- BWH: Andriy Fedorov, Dominik Meier
- UIowa: Hans Johnson
- GE: Jim Miller
- MGH: Greg Sharp
Brachytherapy is a form of radiotherapy, when radioactive sources are permanently implanted into the affected organ. One of the challenges of image-guided brachytherapy is that CT allows good localization of implants, while MRI provides good visualization of the anatomy. The ability to fuse CT and MRI allows accurate dose calculation for the prostate gland and nearby organs that must be spared (e.g., rectum and neurovascular bundles).
We would like to collect ideas and evaluate available solutions for aligning CT and MRI for prostate brachytherapy dose calculation.
Current approaches used in clinic require operator-guided identification of the corresponding seeds followed by point-based registration (we would like to confirm this too with the experienced investigators).
The following approaches to solving this registration problem can be considered:
- identify centroid points manually for the corresponding points, followed by point-based registration (ICP?)
- segment the seeds, and register the surfaces of the segmented labels
- what is the best approach to segment the seeds from CT?
- can we automate seed segmentation from MRI?
- intensity-based registration -- can this be accomplished if we carefully define the ROI for intensity sampling? pre-processing to identify features?
We currently have one clinical dataset that will be used for investigating the considered solutions.
Different approaches to segmentation and registration for this application were discussed with Jim Miller, Luis Ibanez, Tamas Ungi. Initial code for CT fiducial segmentation under development (threshold + connected component analysis + morphological operations to separate the seeds). Tested MR seed enhancement based on grayscale morphological operations. Need to obtain proton density MRI as an alternative modality (reportedly optimal modality for seed imaging).