[OA189] 4D-CBCT streak reduction by simulated undersampling from 3D reconstruction. (August 2018)

• Record Type: Journal Article
• Authors: Jensen, Kasper Rørdam; Bernchou, Uffe; Hansen, Olfred; Brink, Carsten
• Is Part Of: Physica medica. Volume 52(2018)Supplement 1
• Page Start: 72
• Page End: 73
• Publication Date: 2018-08
• Subjects: Medical physics -- Periodicals; Biophysics -- Periodicals; Biophysics -- Periodicals; Imagerie médicale -- Périodiques; Radiothérapie -- Périodiques; Rayons X -- Sécurité -- Mesures -- Périodiques; Physique -- Périodiques; Médecine -- Périodiques; 610.153
• DOI: 10.1016/j.ejmp.2018.06.261
• Languages: English
• ISSNs: 1120-1797

Abstract

Abstract : Purpose: 4D-Cone Beam CT (4D-CBCT) scans used for image guided radiotherapy (IGRT) are usually undersampled and therefore affected by streaking artifacts. This study investigates a streak reduction method that simulates undersampling from a 3D reconstruction. Methods: This study is based on small field of view (SFOV) 4D-CBCT and 4D-CT scans of 20 lung cancer patients subjected to curatively intended IGRT. The proposed method resembles elements of the McKinnon-Bates algorithm [IEEE Trans Biomed Eng., 28, 123, 1981]. CBCT projections were 3D reconstructed and re-projected onto the original gantry angles. Re-projections were reconstructed as 3D and as 4D by respiratory phase binning. Streaks produced due to undersampling were estimated by subtracting the 3D volume from the 4D volumes. Corrected 4D-CBCT was made by subtracting the estimated streaks from a standard 4D-CBCT image. Quantitative analyses were conducted on the expiration phase. Streak-reduction was measured as the ratio of summed gradients in corrected image to uncorrected image, with less than unity indicating improvement. Corrected and uncorrected CBCT images were mapped onto CT by image registration and for each CBCT image the average HU error between CT and CBCT was calculated. Results: The total process took less than 5 min for a 10 phase reconstruction. Visual improvements were observed when applying streak correction for the 4D-CBCT images. The ratio of summed gradients also favored the corrected images (mean = 0.8, SD = 0.05) and no sign of overall degradation of HU values were observed, given that average HU error for corrected CBCT was lower for corrected CBCT (mean = 181, SD = 22.1) than for uncorrected CBCT (mean = 193, SD = 24.7). Conclusions: The quality of SFOV 4D-CBCT images subjected to streak correction was found to be improved based on visual and quantitative validations. A GPU-based implementation could potentially reduce the computation process to less than a minute and thereby making it usable for on-line IGRT.