A Local Evaluation of 2D/3D Registration of an Orthogonal Pair of Kilo Voltage Images Versus Cone Beam CT for Partial Brain Radiotherapy Patients

Abstract

Purpose/Aim

To evaluate two dimensional (2D/2D) kilo voltage imaging platforms, one two dimensional (2D/3D) registration imaging method and CBCT in the setup reproducibility of the partial brain population in a radiotherapy clinic. This research is for the purpose of potentially replacing CBCT with 2D/3D KV reconstruction for a cohort of partial brain radiotherapy patients.

Methods/Process

The comparison of CBCT, 2D2D and 2D3D with the respective images was performed on a Truebeam linear accelerator. This data was transferred to MOSAIQ which is the internal EMR, (electronic medical record). Shifts were extracted from Mosaiq and manually entered into an excel spreadsheet. Patient sensitive information was anonymized within HIPPA (Health Insurance Portability and accountability act, 1996) compliance. After comparing 2D/2D, 2D/3D and CBCT with their respective reference images, shifts in the lateral, longitudinal and vertical direction were noted on all these 3 comparisons. Additionally, the yaw, pitch and roll were noted in the 2D/3D and CBCT comparisons and the mean and standard deviation calculated. This study analysed 177 data samples for each of the three registration methodologies adding to a total of 531 imaging analysed.

Results or Benefits/Challenges

The Wilcoxon signed- rank test was utilized to compare the 3 registration methods. A P-value <0.05 was considered to be significant. The residual error for the 2D3D in the VRT, LAT and LNG direction is -0.0032 cm ± 0.0456 cm, 0.0131 cm ± 0.0441 cm and 0.0392 cm ± 0.0805 cm while for the 2D2D the residual error is found to be respectively -0.0549 cm ± 0.1828 cm, 0.0935 cm ± 0.254 cm and -0.0165 cm ± 0.2224 cm. The average difference in the RTN for the 2D3D versus CBCT was -0.3008° and for the ROLL was 0.0571°, both < 1°. The difference in the ROLL was not found to be statistically significant (p = 0.5672), while the difference in RTN was statistically significant (p = 0.0439) which can be accounted for in the PITCH.

Conclusions/Impact

Although limited by sample size this data collection indicates that the Varian OBI (on board imaging) system on a Truebeam LINAC offers a 2D3D KV image registration tool that displays accuracy levels for target localization that are similar to 3D3D image registration with CBCT and has the potential to replace CBCT for a selected cohort of a partial brain population.