Premium
Technical Note: Determination of the optimized image processing and template matching techniques for a patient intrafraction motion monitoring system
Author(s) -
Tachibana Hidenobu,
Uchida Yukihiro,
Shiizuka Hisao
Publication year - 2012
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.3675404
Subject(s) - artificial intelligence , template matching , pixel , computer vision , computer science , imaging phantom , standard deviation , image resolution , rgb color model , motion detection , cross correlation , tracking (education) , pattern recognition (psychology) , mathematics , nuclear medicine , motion (physics) , image (mathematics) , statistics , medicine , psychology , pedagogy
Purpose: In this work, the authors determine the optimal template matching method and selection of pixel data for use in a system for monitoring patient intrafraction motion. Methods: The motion monitoring system is based on optical tracking of a marker block placed on the patient. The temporal resolution of the system was evaluated with a respiratory motion phantom. The phantom moved the marker with a peak‐to‐peak amplitude of 0.6–4.0 cm and a period of 1, 3, and 6 s. Three template matching methods were applied: Sum of squared difference (SSD), sum of absolute difference (SAD), and normalized cross‐correlation (NCC) using each of four pixel color data schemes (RGB and gray level modified by one of three image processing steps). An in‐house algorithm called auto region‐of‐interest (AutoROI) automatically reset the marker detection region‐of‐interest to improve the calculation speed. Results: RGB and gray level temporal resolutions were 54.22 ± 10.81 (1 SD) s and 12.70 ± 3.87 (1 SD) s, respectively. The temporal resolution when using SSD and SAD was higher than when using NCC. Positional accuracy was within 1 mm. Both values were within the tolerance specified by AAPM Task Group 142. To avoid misidentification of the marker, a threshold‐based self‐validation within the marker recognition system was implemented and was found to improve the tracking of motion with a high amplitude and short period. Conclusions: An intrafraction motion monitoring system using SSD or SAD and applied to gray pixel data can achieve high temporal resolution and positional accuracy.