Experimental validation of Fastcat kV and MV cone beam CT (CBCT) simulator

TL;DR

This paper validates the Fastcat CBCT simulator against real kV and MV CBCT images from a Varian Truebeam linac, demonstrating its accuracy and speed for clinical imaging simulations.

Contribution

It provides the first experimental validation of the Fastcat CBCT simulator with detailed modeling and comparison to actual linac images, confirming its accuracy and efficiency.

Findings

Contrast within 14 and 9 HUs for kV and MV simulations

Detector MTF within 4.2% and 2.5%

Dose within 2.4% and 1.6% of MC calculations

Abstract

Purpose: To experimentally validate the Fastcat cone beam CT (CBCT) simulator against kV and MV CBCT images acquired with a Varian Truebeam linac. Methods: kV and MV CBCT images of a Catphan 504 phantom were acquired using a 100 kVp beam with the on-board imager (OBI) and a 6 MV treatment beam with the electronic portal imaging device (EPID), respectively. The kV Fastcat simulation was performed using detailed models of the x-ray source, bowtie filter, a high resolution voxelized virtual Catphan phantom, anti-scatter grid, and the CsI scintillating detector. Likewise, an MV Fastcat CBCT was simulated with detailed models for the beam energy spectrum, flattening filter, a high resolution voxelized virtual Catphan phantom, and the GOS scintillating detector. Experimental and simulated CBCT images of the phantom were compared with respect to HU values, contrast to noise ratio (CNR),and dose linearity. Detector modulation transfer function (MTF) for the two detectors were also experimentally validated. Fastcat's dose calculations were compared to MC dose calculations performed with Topas. Results: For the kV and MV simulations, respectively: Contrast agreed within 14 and 9 HUs and detector MTF agreed within 4.2% and 2.5%. Likewise, CNR had a root mean squared error (RMSE) of 2.6% and 1.4%. Dose agreed within 2.4% and 1.6% of MC values. The kV and MV CBCT images took 71 and 72 seconds to simulate in Fastcat with 887 and 493 projections, respectively. Conclusions: We present a multi energy experimental validation of a fast and accurate CBCT simulator against a commercial linac. The simulator is open source and all models found in this work can be downloaded from https://github.com/jerichooconnell/fastcat.git