MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Use of Compressed Sensing Optimization for Creation of 3D Printer Friendly Compensators for Use in Small Animal IMRT

X Liu*, G Redler, E Pearson, R D Wiersma, The University of Chicago, Chicago, IL

Presentations

(Tuesday, 7/16/2019) 3:45 PM - 4:15 PM

Room: Exhibit Hall | Forum 4

Purpose: There is significant interest in the development of highly conformal small animal radiation techniques. One method is to perform intensity modulated radiation therapy (IMRT) through the use of 3D printed compensators. This avoids the use of a multi-leaf collimator (MLC), and can be implemented inexpensively and compactly in the limited space of a small animal irradiator enclosure. However, regular IMRT optimization can lead to highly modulated compensator patterns that can be difficult to print due to the limitations of FDM printers. In this work, we present a compressed sensing IMRT optimization approach that reduces compensator modulation without significantly reducing overall dose conformity.

Methods: Compressed sensing (CS) based inverse planning is a method to reduce optimized fluence map complexity. An L1 norm gradient operator can be defined on nearby beamlet intensities in both column-wise and row-wise directions. It is a mixed L1/L2 norm optimization problem where the objective function is non-differentiable and convex, and can be solved efficiently by proximal operator graph solvers. Adding the gradient operator to the objective function, and tuning the weight for the gradient operator, smoother fluence maps can be obtained.

Results: To evaluate the approach, several C-shape TG199-like phantoms were generated, and different objective functions where dose deviation/underdose/overdose quadratic function were considered. In most cases, the summation of the gradients of all fields was reduced 10-30%, while DVH curves remained acceptable.

Conclusion: The work uses CS in IMRT to reduce fluence map complexity for easy 3D printer compensator creation. Preliminary results show that CS can produce significantly smoother fluence maps such that converting the fluence map into physical compensator thickness maps results in compensators with smoother, more easily printable surfaces.

Keywords

Compensators, Inverse Planning, Optimization

Taxonomy

TH- Small Animal RT: Planning

Contact Email