Projected digital radiologic images for teaching: balance of image quality with data size constraints

The authors performed this study to determine, in the context of a teaching presentation with slides, the information content needed in a digitized radiologic image for it to be equivalent in quality to an analog image.

Eleven sets of radiologic images were obtained from a teaching file collection and digitized. The images were scanned at high resolution and saved as image files. The information content of each image was then halved repeatedly with repeat sampling of the image. This procedure was repeated nine times to yield a total of 10 images with an information content that ranged from very low (32 kB) to high (12 MB). Each of these image files was made into a 35-mm slide by using a digital slide maker. The original radiographs were subsequently made into slides by using conventional photographic methods. Care was taken to make the images identical in every respect except digitization and information content. The slides were shown to radiologists, who filled out questionnaires to rate image quality.

Digitized slides that contain an average of 416 kB of data (the equivalent of a 750 x 570-pixel matrix with 8-bit gray levels) were equivalent to analog images for 90% of viewers. At 830 kB (1,060 x 800-pixel matrix with 8-bit gray levels), 100% of viewers found no difference between digitized and analog images.

Digital images become indistinguishable from analog images if a sufficient amount of information is retained in the digital image. There is a defined relationship between the perception of quality and the information content of a digital image.