Comparison of two methods for evaluating image quality of chest radiographs

A set of 15 analog chest images was digitized with a high performance scanner and manipulated in terms of noise and resolution to yield three sets of images; the original plus two with different noise and resolution properties. These sets were evaluated with Visual Grading Analysis (VGA) where the observer rates the visibility of certain normal anatomical structures (as described by the European Quality Criteria) compared to a reference image. One of the non-manipulated digitized images was used in an ROC-related method – the Free response Forced Error experiment. Simulated lesions of different contrast and size were randomly superimposed on 50 copies of this image. The images were then manipulated in the same way as the images used for VGA. All observations were done by a group of seven expert radiologists from six different European countries and all images were printed back to film before evaluation. The ranking of the image manipulations was the same for the two methods. Although a strong correlation cannot be predicted with only three sets of images, it is encouraging that the simpler VGA study in some cases might be used as a replacement for the more time- and effort-consuming ROC analysis