Investigation of the Water Equivalent Depth Method for Rapid Evaluation of the Dosimetric Accuracy of Synthetic CT
J Choi1*, B Asadi2, J Welsh1, S Chalup1, J Dowling1; 3, J Simpson2, P Greer1; 2, (1) University of Newcastle, NSW, AU, (2) Calvary Mater Newcastle, NSW, AU, (3) CSIRO, QLD, AU,
Presentations
PO-GePV-M-251 (Sunday, 7/10/2022) [Eastern Time (GMT-4)]
ePoster Forums
Purpose: To investigate how image similarity metrics and the water equivalent depth (WED) difference calculation are correlated to the absorbed radiation dose agreement between synthetic CT (sCT) scan and reference CT. To derive a simple but dosimetrically meaningful assessment of sCT accuracy for both model generation and assessment.
Methods: Retrospective CT and atlas-based synthetic CT of 62 (53 IMRT and 9 VMAT) prostate cancer patients were used. For image similarity measurements, the sCT and the CT (reference) were aligned using clinical registration parameters. The MAE and ME in HU, PSNR, SSIM between the two scans were measured. The mean WED error measurements were automatically performed from the skin surface to the beam isocenter at 36 gantry angles. The image similarity measures and WED difference calculations were then compared to the isocenter dose agreement between the sCT and the reference CT scans.
Results: While no particular relationship to dose was observed for the other image similarity metrics, the ME results showed a linear trend against the isocenter dose difference. The R-squared value of the trend was 0.6, and its 95 % prediction interval span covers between -1.1 % and 1.2 % isocenter dose difference. The mean WED error results showed a linear trend (R² = 0.8) with the narrower 95 % prediction interval span from -0.8 % to 0.8 % isocenter dose difference.
Conclusion: Our results showed that the most commonly used image similarity measure in sCT development and assessment (MAE) has no relationship to the dosimetric utility of the scan. The mean WED error between the sCT and the reference CT highly correlates with the difference in isocenter dose between the two scans. Therefore, it can facilitate the process of developing and evaluating new sCT generation algorithms if used in conjunction with the other image representation similarity metrics.
Funding Support, Disclosures, and Conflict of Interest: This research was supported by an Australian Government Research Training Program (RTP) Scholarship and the Australian MRI Linac Research Program Grant
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email