Click here to


Are you sure ?

Yes, do it No, cancel

A Fast Dose and Fluence Optimization Engine Using Matrix Sparsity for Adaptive Radiotherapy

H Stephens1*, Q Wu1, W Wang1, X Li1, J Zhang1, C Wang1, Y Sheng1, Y Ge2, Q Wu1, (1) Duke University, Durham, NC, (2) University of North Carolina at Charlotte, Charlotte, NC,


(Sunday, 7/12/2020)   [Eastern Time (GMT-4)]

Room: AAPM ePoster Library

Purpose: To develop an accurate and fast dose and fluence optimization algorithm for the purposes of adaptive radiotherapy.

Methods: Using a Finite-Sized Pencil Beam model designed for IMRT optimization, the dose distribution is calculated to a homogenous water volume. Defining the voxels of interest as any voxels within a structure with a dose-volume constraint, the voxels are then mapped to the water volume by projecting onto the beam basis and calculating a radiological depth for heterogeneity corrections. For each structure, a beamlet influence matrix is calculated using an analytical model of the beamlet dose, a beamlet’s interaction neighborhood, and the dose calculated from a uniform fluence map. Then, using sparse matrix algebra the fluence is optimized in a rapid fashion using an analytical description of the gradient.

Results: Optimizing a plan consisting of 7 beams on a prostate case, the total computation time from the beginning of the dose calculation through all the optimization took 61.07 s. This included the 3D dose calculation, beamlet dose calculations, and the fluence optimization. Comparing our calculated fluence maps to those calculated using Eclipse, we found an average normalized mutual information index of around 0.6. Importing our optimal fluence into Eclipse we found a difference in mean dose for the PTV, Rectum, and Bladder to be 0.4%, 0.9%, and 2.6% and an average difference in the DVH of -0.38%, -0.85%, and 2.43% of volume at dose.

Conclusion: Our algorithm showed very solid results in producing a plan comparable to the plan produced by Eclipse. We observed better results with some structures, such as the bladder, where we exceeded the goals for optimization. These positive results in conjunction with the relative size and speed proved its efficacy in adaptive applications.


Not Applicable / None Entered.


TH- External Beam- Photons: IMRT/VMAT dose optimization algorithms

Contact Email