Projection image compression method for on-board radiotherapy imaging system
10.3969/j.issn.1005-202X.2025.10.005
- VernacularTitle:放疗机载成像系统的投影图像压缩方法
- Author:
Zhixing CHANG
1
;
Jiawen SHANG
1
;
Yuhan FAN
1
;
Jianrong DAI
1
;
Hui YAN
1
Author Information
1. 国家癌症中心/国家肿瘤临床医学研究中心/中国医学科学院北京协和医学院肿瘤医院放射治疗科,北京 100021
- Publication Type:Journal Article
- Keywords:
cone-beam projection image;
radiotherapy;
image compression;
dynamic video encoding algorithm;
static image coding algorithm
- From:
Chinese Journal of Medical Physics
2025;42(10):1289-1297
- CountryChina
- Language:Chinese
-
Abstract:
Objective To achieve efficient compression of on-board radiotherapy projection images using dynamic video encoding algorithms.Methods The on-board radiotherapy imaging system primarily provides 2D X-ray projection images for patient positioning verification and 3D tomographic image reconstruction.Since multiple projection images acquired continuously exhibit strong spatiotemporal correlations,their similarities could be used to eliminate redundant information,thereby improving the image compression ratio.During image compression,the image sets obtained at different times were arranged into an image sequence which was input into a video encoder and output as a video file.During image decompression,the video file was input into a video decoder and output as an image sequence,and the images in the sequence were then assigned back to their original image sets.Three current dynamic video encoding algorithms(AVC,HEVC,and AV1)and the classic static image coding algorithm(JPEG 2000)were tested on a database of 2D projection images.The performance of various compression algorithms was evaluated using indicators such as compression ratio(CR),peak signal-to-noise ratio(PSNR),and structural similarity(SSIM).Moreover,visual comparison of projection images before and after compression was evaluated by clinical radiation oncologists.Results Dynamic video encoding algorithms achieved higher CR than the static image coding algorithm.The average CR of the 3 dynamic video encoding algorithms was as followed:CRAVC=11.50,CRHEVC=30.74,and CRAV1=27.10,while the average CR of the static image coding algorithm(JPEG 2000)was 5.28.For abdominal projection images,well-defined contours and textural details were preserved even when the CR reached 42.37.For head-neck projection images,although mild contour blurring emerged at a CR of 20.71,subsequent evaluation by clinical radiation oncologists confirmed that the reconstructed CBCT images still satisfied clinical requirements.Conclusion These dynamic video encoding algorithms effectively utilize the strong correlation information between multiple projection images,reduce the storage of redundant information,and greatly improve the image CR.
