Access Type

Open Access Dissertation

Date of Award

January 2017

Degree Type


Degree Name



Medical Physics

First Advisor

Jay Burmeister


Purpose: With a fast adoption of emerging technologies, it is critical to fully test and understand its limits and capabilities. In this work we investigate new graphic processing unit (GPU) based treatment planning algorithm and its applications in helical tomotherapy dose delivery. We explore the limits of the system by applying it to challenging clinical cases of total marrow irradiation (TMI) and stereotactic radiosurgery (SRS). We also analyze the feasibility of alternative fractionation schemes for total body irradiation (TBI) and TMI based on reported historical data on lung dose and interstitial pneumonitis (IP) incidence rates.

Methods and Materials: An anthropomorphic phantom was used to create TMI plans using the new GPU based treatment planning system and the existing CPU cluster based system. Optimization parameters were selected based on clinically used values for field width, modulation factor and pitch. Treatment plans were also created on Eclipse treatment planning system (Varian Medical Systems Inc, Palo Alto, CA) using volumetric modulated arc therapy (VMAT) for dose delivery on IX treatment unit. The constraints ware selected to ensure that at least 95% of the PTV received the prescription dose while minimizing the doses to OARs which consisted of lungs, heart, liver, kidneys, brain, and small bowel. Resulting plans were evaluated based on plan quality, optimization and dose calculation times, and beam on times. Gamma indices (Γ) were also used to compare planar dose distributions between the planning systems. In addition a plan quality index (Q) was developed for quantitative analysis of relative plan quality which included mean and maximum doses. The GPU planning systems was also evaluated for single fraction radiosurgery/SBRT capabilities. Treatment plans were created for spine metastases based on national protocol RTOG 0631 and the dosimetric results were compared to four other modalities.

A retrospective review was performed of 42 publications that reported IP rates along with lung dose, fractionation regimen, dose rate and chemotherapy. The analysis consisted of nearly thirty two hundred patients and 34 unique radiation regimens. Multivariate logistic regression was performed to determine parameters associated with IP and establish does response function.

Results: The results showed very good dosimetric agreement between the GPU and CPU calculated plans. A gamma analysis Γ(3%, 3 mm) < 1 of the GPU plan resulted in average of 97% of calculated voxels satisfying Γ < 1 criterion as compared to baseline CPU plans. The optimization/dose calculation time with the new GPU system is about 20 times faster than with the CPU system. Is was also about 4 times faster than Eclipse treatment planning system while achieving superior OAR dose sparing ranging from 3% to 52%. Analysis of optimization parameters showed increase in plan quality index (Q) with lower pitch, smaller field size and higher modulation factor. The beam on time increases with increasing plan quality index and associated optimization parameters with the highest effect observed with field size.

The results from SBRT study show that GPU planning system can maintain 90% target coverage while meeting all the constraints of RTOG 0631 protocol. Beam on time for Tomotherapy and flattening filter free RapidArc was much faster than for Vero or Cyberknife.

Retrospective data analysis showed that lung dose and Cyclophosphomide (Cy) are both predictors of IP in TBI/TMI treatments. The dose rate was not found to be an independent risk factor for IP. The model failed to establish accurate dose response function, but the discrete data indicated a radiation dose threshold of 7.6Gy (EQD2_repair) and 120 mg/kg of Cy below which no IP cases were reported.

Conclusion: The TomoTherapy GPU based dose engine is capable of calculating TMI treatment plans with plan quality nearly identical to plans calculated using the traditional CPU/cluster based system, while significantly reducing the time required for optimization and dose calculation. The new system was able to achieve more uniform dose distribution throughout the target volume and steeper dose fall off, resulting in superior OAR sparing when compared to Eclipse treatment planning system for VMAT delivery. The machine optimization parameters tested for TMI cases provide a comprehensive overview of the capabilities of the treatment planning station and associated helical delivery system. The new system also proved to be dosimetrically compatible with other leading modalities for treatments of small and complicated target volumes and was even superior when treatment delivery times were compared. These finding demonstrate that the advanced treatment planning and delivery system from TomoTherapy is well suitable for treatments of complicated cases such as TMI and SRS and it’s often dosimetrically and/or logistically superior to other modalities. The new planning system can easily meet the constraint of threshold lung dose established in this study. The results presented here on the capabilities of Tomotherapy and on the identified lung dose threshold provide an opportunity to explore alternative fractionation schemes without sacrificing target coverage or lung toxicity.