Ablative radiation doses in inoperable locally advanced pancreatic cancer (LAPC) delivered millimeters from the gastro-intestinal (GI) tract have been shown to produce a survival duration similar to resection in operable tumors. The safe delivery of these doses requires solutions that control or account for motion of the GI tract. Most importantly, the delivered dose to mobile organs at risk (OARs) must be verified, but currently there is a lack of appropriate software tools for its accurate quantification in cone-beam computed tomography (CBCT) images at treatment. The long-term goal is to increase the potential for long-term survival in patients with LAPC by enabling more widespread use of ablative dose delivery near the GI tract. The objective of this application is to develop and evaluate the feasibility of fast CT-to-CBCT deformable registration algorithms for determining delivered dose to OARs in these types of treatments. The central hypothesis is that the proposed methodology can accurately compute organ deformations from pretreatment CBCT scans rapidly enough to ensure the feasibility of the computation and assessment of daily delivered dose immediately prior treatment. This hypothesis is based on preliminary data which indicate that OAR proximity to the target can vary considerably over interfractional time scales. The rationale is that completion of this research will provide a means of quantitatively determining delivered dose to OAR, thereby facilitating ablative dose delivery to nearby targets.
Our specific aims are: 1) develop new algorithms to determine the dose delivered to OARs from radiation treatment in the pancreas; and 2) evaluate the accuracy of the algorithms in retrospective analysis of patient images from ablative dose treatments in the pancreas. In the first aim, we will develop a fast image registration approach based on our recent work on fast predictive multimodal image registration that is customized for CT-to-CBCT alignment using existing patient image data, and evaluate accuracy by using an independent data set to compare predicted OAR segmentations in the CBCT images to controls consisting of manual segmentations by experts. In the second aim, we will evaluate the predicted deformable registrations using the distance discordance metric in a statistical sampling technique, and compute delivered dose to assess its deviation from the planned dose. The approach is innovative because it introduces novel algorithms to overcome the obstacles in achieving fast and accurate deformable CT-to-CBCT registrations for guiding radiotherapy of pancreatic cancer. This contribution is significant because the methodology, when integrated into a clinical workflow, will help to control the variability of delivered dose to OARs more accurately than the current practice of qualitative assessment of OAR positions in CBCT-guided treatments. The proposed research is the first step towards facilitating more widespread use of ablative dose delivery near the GI tract, thereby offering the potential for long-term survival to larger numbers of patients with LAPC.
In inoperable locally advanced pancreatic cancer, radiation treatment with ablative doses has been shown to produce survival durations that are comparable to surgical removal of operable tumors. The safe delivery of ablative doses within millimeters of the gastro- intestinal (GI) tract requires accurate knowledge of the cumulative dose to organs at risk that move day to day over a course of treatment. The proposed research is relevant to public health because it will make software tools available that calculate the cumulative dose to organs at risk, making possible the safe delivery of ablative doses to tumors near the GI tract by clinicians in the community at large.
