Control portal images enhancement

The general aim of our study is the patient setup verification using control portal images during the treatments. The main problem of the control portal images processing is their poor quality. We investigated a way of improving the image quality in order to allow a segmentation stage. Four items were studied for this purpose: the presence of distortion, the background gradient of intensity correction, the noise reduction, and the image restoration. The absence of significant distortion was checked with a grid pattern phantom. The background correction was performed by the subtraction from the original image of a smoothed version of this image. We tested 14 noise reduction methods. The most appropriate for our images is the truncated average. Four restoration techniques were compared. The Maximum a Posteriori algorithm is the most efficient. For all the imaging systems, the energies, and the anatomical zones tested an important improvement in anatomic detail was apparent. For example, the SNR of a SRI-100 pelvis image (4 monitor units at 10 MV), went up from 0.97 to 20.06 after enhancement. The image quality improvement will facilitate or enable the segmentation of the control portal images. It is the first step in the patient setup verification.