MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

A Multi-Institutional Feasibility Study of a Novel Cloud-Based and GPU-Accelerated Clinical Workflow for Adaptive Radiotherapy

A Santhanam1 , M Cao1*, P Potrebko2,3 , D Elliott2 , M Lauria1 ,S Seshan2 , R Bosley2 , D Low1 , (1) University of California, Los Angeles, Los Angeles, CA, (2) UCF School of Medicine, Orlando FL, (3) SegAna Inc, Orlando, FL

Presentations

(Wednesday, 7/17/2019) 10:00 AM - 10:30 AM

Room: Exhibit Hall | Forum 5

Purpose: There is a clinical need to systematically investigate adaptive radiotherapy (ART) using on-board imaging modalities. Specifically, the tasks related to adaptive radiotherapy need to be automated in order to enable a fast and easy clinical workflow, where patient anatomical changes and setup uncertainties can be accounted for in order to compute and optimize the delivered dose.

Methods: We investigated a novel cloud-based and GPU-accelerated software workflow (RTApp, SegAna LLC) for analyzing on-line and off-line variants of ART. A set of 20 patient datasets (5 SBRT, 5 IMRT, 5 Conformal, and 5 VMAT) were used for this feasibility study. Each patient dataset consisted of the planning kVCT, daily CBCTs with their corresponding daily shifts in RTReg format, treatment plan and contoured structures. The software presented an automated workflow for (a) retrieving planning and daily image datasets from a clinical database, (b) performing deformable image registration between the planning and the daily image using a well-validated algorithm, (c) propagating planning contours to the current on-board imaging, (d) deforming and overlapping the planning dose in order to compute the predicted dose to be delivered, and (e) generating dose for each of the contoured structures based on the on-board imaging anatomy. The software was designed to exploit the GPU capabilities and is hosted on a cloud-based computing framework.

Results: The data was processed on an Azure computing setup with an Nvidia GPU dedicated for the computations. The total computation time required for each fraction was approximately 3 minutes and was observed to be dependent on the number of contour substructures to be processed.

Conclusion: In our evaluation, the automated workflow was observed to be computationally capable of supporting the needs of adaptive radiotherapy. The framework can help clinicians to develop and customize specific workflows and dosimetric endpoints to facilitate ART.

Funding Support, Disclosures, and Conflict of Interest: This work was funded by the National Science Foundation SBIR Phase I award. Financial Disclosure: Anand Santhanam is also the founder of SegAna and so has vested interest in it.

Keywords

Not Applicable / None Entered.

Taxonomy

Not Applicable / None Entered.

Contact Email