Loss of diagnostic information in digital images.

Abstract

Versteeg and her colleagues, in their excellent recent article, report that a reduction in the size of digital images leads to a loss of detectability of small endodontic ®les. In this context they mean by a reduction in size that fewer pixels are displayed on the monitor than are present in the sensor. They base their study on the Visualix (Gendex, Monza, Italy) with the latest generation CCD sensor, in conjunction with the Emago Advanced (Acta, Amsterdam, The Netherlands) software. The authors have assumed that de-zooming (zooming out) as a means of diminishing the image size displayed automatically involves trashing (clipping) a (large) fraction of the pixels from the original image, thus leading to a loss of information. This is, in fact, the method used by the vast majority of the image processing software currently available and, apparently, by the Emago Advanced system. At Gendex Dental Systems, Milan, we have been aware of this potential problem since our initial involvement with digital imaging for dental radiography. The ®rst version of Visualix in 1992 (sometimes referred to as Visualix/VIXA 1 or 2) operated in a DOS environment. This meant that there was a one-to-one correspondence between the pixels in the sensor, the memory and the monitor which was therefore limited to operating in one speci®c video mode. This approach is no longer practical with the current Windows-based systems. Gendex now equips all its digital imaging products, including the Visualix system used by Versteeg et al with the Windows-based VixWin software (currently version 2.4). We have devised a special technique (details of which remain con®dential) in order to minimise the loss of diagnostic information when an image is dezoomed (or down-sized) for display on a monitor. The proprietary display algorithm used in this case does not function in the straightforward way, leading to the loss of information assumed by Versteeg et al. In the circumstances, I would con®dently anticipate that the outcome of a similar study, using the currently-available VixWin 2.4 software instead of Emago, would have been better detection of small endodontic ®les in scaled-down images.