Click here to


Are you sure ?

Yes, do it No, cancel

Novel Wiener-Based Digital Filtering Significantly Improves Intrafraction Kilovoltage Image Quality

J Sun1*, P Keall1 , J Booth2 , T Eade2 , D T Nguyen1 , (1) ACRF Image-X Institute, University of Sydney, Camperdown, NSW, (2) Northern Sydney Cancer Center, Royal North Shore Hospital, St Leonards, NSW


(Thursday, 8/2/2018) 10:00 AM - 12:00 PM

Room: Room 207

Purpose: This work investigates the utility of a temporal-spatial statistical filter algorithm to improve intrafraction kV images in the presence of MV scatter noise on phantom and patients with different hip-widths.

Methods: X-ray images (10fps, 125kVp, 80mA and 13ms) of an anthropomorphic phantom (Rando) with implanted fiducials were acquired using the kV imaging system of a Varian Trilogy linac. The hip-width of the phantom was 35.5 cm, which was increased to 39.5cm and 43.5cm by adding blocks of 2cm solid water laterally. For each phantom size, images were taken with no MV beam on, MV beam at 300MU/min and 600MU/min. The algorithm was also tested on lateral intrafraction kV images of 4 patients with large hip-width (38-40cm) . Our filtering method consists of a temporal filter, a spatial Gaussian filter and finally a statistical Wiener filter. Contrast-to-Noise ratios (CNRs) were calculated to assess the effect of the proposed filter, in comparison with temporal frame averaging filter (3 frames) currently used in the Kilovoltage Intrafraction Monitoring (KIM) system.

Results: For all phantom and patient images with MV scatter, our method improved CNRs compared with temporal averaging. For the 43.5cm phantom, our method improved the CNRs from 0.2±0.2 to 1.0±0.7 (p<0.01) for the 600MU/min scatter and from 0.2±0.1 to 1.0±0.5 (p<0.01) for the 300MU/min. Without scatter, our method improves the CNR of the 43.5cm phantom from 0.7±0.1 to 1.8±0.3 (p<0.01). For patients images, the CNRs on average were improved from 0.49±0.53 to 1.54±1.46 (p<0.01).

Conclusion: A Wiener-based filtering method was developed to mitigate the effects of MV scatter on intrafraction kV images, with superior performance compared with frame averaging. The work implications is that kV images of larger patients can be used for intrafraction monitoring purposes while imaging parameters can be lowered for smaller patients, thus reducing additional imaging dose.

Funding Support, Disclosures, and Conflict of Interest: D T Nguyen is funded by an NHMRC and Cancer Institute New South Wales Early Career Fellowship. P Keall is funded by an NHMRC Senior Principal Research Fellowship.


Not Applicable / None Entered.


IM- X-ray: Image processing

Contact Email