Development and clinical implementation of an enhanced display algorithm for use in networked electronic portal imaging.
Van den Heuvel F., Han I., Chungbin S., Strowbridge A., Tekyi-Mensah S., Ragan DP.
PURPOSE: To introduce and clinically validate a preprocessing algorithm that allows clinical images from an electronic portal imaging device (EPID) to be displayed on any computer monitor, without loss of clinical usability. The introduction of such a system frees EPI systems from the constraints of fixed viewing workstations and increases mobility of the images in a department. METHODS AND MATERIALS: The preprocessing algorithm, together with its variable parameters is introduced. Clinically, the algorithm is tested using an observer study of 316 EPID images of the pelvic region in the framework of treatment of carcinoma of the cervix and endometrium. Both anterior-posterior (AP/PA) and latero-lateral (LAT) images were used. The images scored were taken from six different patients, five of whom were obese, female, and postmenopausal. The result is tentatively compared with results from other groups. The scoring system, based on the number of visible landmarks in the port, is proposed and validated. Validation was performed by having the observer panel score images with artificially induced noise levels. A comparative study was undertaken with a standard automatic window and leveling display technique. Finally, some case studies using different image sites and EPI detectors are presented. RESULTS: The image quality for all images in this study was deemed to be clinically useful (mean score >1). Most of the images received a score which was second highest (AP/PA landmarks > or =6 and LAT landmarks > or =5). Obesity, which has been an important factor determining the image quality, was not seen to be a factor here. Compared to standard techniques a highly significant improvement was determined with regard to clinical usefulness. The algorithm performs fast (less than 9 seconds) and needs no additional user interaction in most of the cases. The algorithm works well on both direct detection portal imagers and camera-based imagers whether analog or digital cameras. CONCLUSIONS: We have demonstrated that it is possible to preprocess EPIs in such a way that the clinically relevant landmarks are easily detected on a generic computer screen. The algorithm is system-independent and fast. This allows for the encoding of EPIs in more generalized commercial formats so that distribution of images is facilitated.