To compare the diagnostic performance of linear and logarithmically contrast-enhanced subtraction images, acquired with digital and digitized radiographs, in detecting approximal enamel subsurface demineralization. Fifty caries-free human third molars were immersed in a demineralizing solution for 60, 75, 90, and 120 days, in order to induce artificial enamel subsurface demineralization. The teeth were coated with nail varnish, leaving only a circular window of approximately 7 mm2 in one of the approximal surfaces, allowing contact with the solution. Standardized radiographs of the teeth were taken prior to and after the demineralization period with three digital systems, CygnusRay MPS, DenOptix and DIGORA, and InSight film. Conventional, digital and digitized radiographs were assessed by three experienced radiologists. Linear and logarithmically contrast-enhanced subtraction images were acquired and then examined by a fourth independent radiologist. For the validation of the radiographic diagnosis, the enamel test areas were submitted to Knoop microhardness profiling. Radiographic interpretation data was evaluated using ROC analysis. The areas under the ROC curves (Az) were compared by the chi-squared test. The level of significance was set at P=0.05. No significant differences were found between linear and logarithmically contrast- enhanced subtraction images, acquired with the four studied modalities: CygnusRay MPSlinear (Az=0.95), CygnusRay MPSlog (Az=0.98), DenOptixlinear (Az=0.97), DenOptixlog (Az=0.99), DIGORAlinear (Az=0.98), DIGORAlog (Az=0.98), digitized radiographylinear (Az=0.99), digitized radiographylog (Az=0.99). Linear and logarithmically contrast-enhanced subtraction images, acquired with digital and digitized radiographs, were diagnostically comparable for assessing enamel subsurface demineralization.34133-9