Digital subtraction radiography for diagnosis of periodontal bone lesions with simulated high-speed systems.

Standardized radiographs were obtained of the molar and premolar areas of dry human mandibles before and after small lesions were induced in the alveolar bone crest. The radiographs were digitized, and various amounts of random noise were added to simulate images produced by more efficient, low-radiation-dose systems. Subtraction images were produced from these noise-degraded radiographs. The diagnostic performance of human observers from these images was then compared with that obtained from both subtraction images made from nondegraded radiographs and from the original, conventional images. The diagnostic accuracy obtained with subtraction images, even if they were made from severely noise-degraded images, was equal or superior to that obtained with the conventional radiographs. This indicates that structured noise, which is reduced by the subtraction procedure, is a predominant, limiting factor in the diagnosis of small periodontal bone lesions. It also suggests that the application of a subtraction technique for the detection of such changes occurring over a period of time can make possible the use of significantly faster imaging systems than those currently employed.