MENU

Click here to

×

Are you sure ?

Yes, do it No, cancel

Automatic Detection of Graticule Isocenter and Scale From KV and MV Images

J Yang1*, R Fang2 , W Du1 , L Court1 , (1) UT MD Anderson Cancer Center, Houston, TX, (2) Rice University, Houston, TX

Presentations

(Tuesday, 7/31/2018) 10:30 AM - 11:00 AM

Room: Exhibit Hall | Forum 5

Purpose: To automate the detection of isocenter and scale of mechanical graticule on kV/MV films or EPID images for applications in low-resource settings.

Methods: We developed a robust image processing approach to automatically detect isocenter and scale of mechanical graticule from digitized kV/MV films and EPID images. A series of preprocessing steps were first applied to the digital images, including correlation with the tick mark template images, linear weighting to rescale the correlation values, and non-maxima suppression to remove spurious axial lines. Then a combination of Hough transform and Radon transform was performed to detect the graticule axes and isocenter. The magnification of the graticule was automatically detected by solving an optimization problem using golden section search and parabolic interpolation algorithm. Tick marks of the graticule were then determined by extending from isocenter along the graticule axes with multiples of the magnification value. This approach was validated using a number of 23 kV films, 26 MV films, and 91 EPID images in different anatomical sites (head-and-neck, thorax, and pelvis). Accuracy was measured by comparing computer detected results with manually selected results.

Results: The proposed approach was robust to kV/MV films of varying image quality. The isocenter was detected with an accuracy of less than 1 mm for 98% images except 3 kV films where graticule was not actually visible. Of all images with correct isocenter detection, 99% had a magnification detection error less than 1% and tick mark detection error less than 1 mm, with the exception of 1 kV film (magnification error: 6.15%; tick mark error: 3.83 mm) and 1 MV film (magnification error: 0.45%; tick mark error: 1.11 mm).

Conclusion: We developed an approach to robustly and automatically detect graticule isocenter and scale from 2D kV/MV films, which allows automated treatment planning based on 2D x-ray images.

Funding Support, Disclosures, and Conflict of Interest: This research was supported in part by an NCI UH2 grant CA202665.

Keywords

Image Processing, Pattern Recognition

Taxonomy

IM/TH- Image Analysis (Single modality or Multi-modality): Image processing

Contact Email