Quality enhancement of Pelvis electronic portal images in order to improve treatment accuracy

Document Type: Conference Proceedings

Authors

1 Department of Biomedical Engineering and Medical Physics Shahid Beheshti University of Medical Sciences, Tehran, Iran

2 Shohadaye haftome_tir hospital, Oncology department, Tehran, Iran

Abstract

Introduction: In order to reduce patient setup error in treatment field, portal images with treatment beam (megavoltage x-ray) are widely used. These images are acquired by electronic portal imaging devices (EPID). However, portal images suffer from insufficient anatomical information, contrast, and spatial resolution, because of the fact that Compton scatter is the main photon interaction in megavoltage range. The aim of this study is to inspect the efficacy of multiresolution processing method on enhancing quality of these images, which improves the accuracy of the setup verification.
Materials and Methods: images of pelvis area were captured which had a low level of quality. Therefore, three major steps are required: contrast enhancement, noise reduction and edge sharpening. A multiresolution processing which can decompose and combine the different scales information are proposed to overcome deficiency of histogram-based enhancement methods. EPID images were decomposed to details and approximation components in order to observe structures in different scales. As a result, the components with highest level of noise and them with bone structure information can be identified. The components can be weighted for reducing the noise and sharpening the edges. Finally modified components are used to reconstruct the enhanced images. Anatomical landmarks in processed images were scored visually by observers in three different levels (physician, physicist and technologist) to qualify the efficacy of the algorithm.
Results: The pelvis anatomical landmarks (sacrum, crest of ilium, symphysis, ischium, pubis, acetabulum, sacroiliac joint) were compared by many experts as a quantitative evaluation in terms of their quality, showing the superiority of the processed images. This improvement was reported by all of the experts in different levels, which was mentioned above.
Conclusion: it was seen that noise and anatomical structures are separable in different components. The results show that the proposed method in this study significantly improved the quality of portal images, which leads to much more accurate organ localization during radiotherapy of pelvis area.

Keywords