Standard edge detection algorithms versus conventional auto-contouring used for a three-dimensional rigid CT-CT matching

Abstract

Background: To reduce uncertainties of patient positioning, the Computerized Tomography (CT) images acquired at the treatment planning time can be compared with those images obtained during radiation dose delivery. This can be followed during dose delivery procedure as Image Guided radiotherapy (IGRT) to verify the prescribed radiation dose delivery to the target as well as to monitor radiation dose constraints for organ at risks located in the vicinity of tumour region. A method was developed to compare registered rigid CT images with those acquired during treatment procedure. Materials and Methods: Several CT images were acquired for a typical Rando phantom at head and neck region. Selecting the CT images as reference, they were then manipulated in transitional and rotational directions. The differences in transited and rotated images were evaluated by edge detection algorithms and conventional automatic contouring used in most of current treatment planning systems. Setting of standard edge detection algorithms was investigated and the appropriate one was selected. Applying the selected optimized standard edge detection algorithm and conventional auto-contouring on CT image differences, the characteristics of methods were evaluated. Results: Results show that 1 pixel difference in transition and 1 degree in rotation can be recognized for inhomogeneous regions. A significant variation was detected at the bony-soft tissue and air-soft tissue conjunction regions. Conclusion: The results obtained from the current study are comparable with those reported using Chamfer algorithm. It is concluded that the current method, can be used to control patient positioning in radiotherapy sessions as a part of Image guided radiotherapy protocols.