par Verellen, Dirk;De Neve, Wilfried;Van den Heuvel, Frank;Coghe, Marc;Storme, Guy ;Louis, Olivia
Référence International journal of radiation oncology, biology, physics, 27, 4, page (945-952)
Publication Publié, 1993-11
Article révisé par les pairs
Résumé : Purpose: A test of several image enhancement techniques, performed on on-line portal images in real clinical circumstances, is presented. In addition a score system enabling us to evaluate image quality on pelvic fields is proposed and validated. Methods and Materials: Localization images (n = 546) generated by an on-line portal imaging system during the treatment of 13 patients on pelvic fields were obtained by delivering a radiation dose of 6-8 cGy by an 18 MV photon beam, and recorded with a silicon intensified target video camera with adjustable gain, kV- and black level. Set-up errors were corrected before continuing irradiation. A scoring system based on the number of visible bone-soft tissue edges and transformed to a scale 0 to 5 was developed to judge image quality. A validation of this classification of images was performed with the use of transsectional bone-densities (bone-density *radiological path length) specified at the score defining landmarks. A high pass filter was used on all images, additional online open field subtraction was performed on 242 fields. Off-line study was performed in which a panel consisting of two groups (one composed of three radiation oncologists, the other of three radiotherapy technologists), scored 470 pelvic fields without further enhancement, and the same images with Contrast Limited Adaptive Histogram Equalization (CLAHE) (Pfizer et al.). Two different clipping levels (3.0 and 5.0) were studied. Results: Gender and transsectional bone-densities were the most defining patient-related factors influencing image quality. Camera settings, gantry angle, and image post-processing were important non-patient-related factors. All investigators judged CLAHE to ameliorate low contrast images and to deteriorate good quality images (p < 0.001). © 1993.