Image deconvolution as an aid to feature identification: A clinical trial

The focus of this paper is to evaluate the clinical performance of the image processing technique which we have developed for computed radiography x-rays. This algorithm, which was presented at the SPIE '99 medical imaging conference, uses iterative deconvolution with a measured point spread function to reduce the effect of scatter. Wavelet denoising is also carried out after each iteration to remove effects due to noise. A random selection of chest x-rays were processed using the algorithm. Both the raw and processed images were presented to the radiologists in a random order. They scored the images with regard to the visibility of anatomical detail and image quality as outlined in the european guidelines on quality criteria for diagnostic radiographic images. The most notable result of the technique is seen in the reduction of noise in the processed image.