The aim of this study was to compare the accuracy of approximal caries lesion depth measurements in radiographs from four digital systems. Two CCD-based sensors, Dixi (Planmeca) and Sidexis (Sirona), and two phosphor plate systems, Digora (Soredex) and DenOptix (Gendex), were used to record radiographs of 177 extracted human teeth. Sixty-four radiographically visible lesions were selected and measured by 4 observers in the digital radiographs and in digital photographs of histological sections of the teeth. The gold standard (true measure) was defined as the mean of the 4 observers’ measurements on the histological sections. Two untrained observers underestimated general lesion depth in all the radiographic systems while each of 2 trained observers underestimated the depth in one of the systems. Analysis of variance revealed that Digora images resulted overall in the smallest underestimation, followed by Dixi, DenOptix and Sidexis. The differences between Digora and Sidexis and between Dixi and Sidexis were statistically significant (p < 0.05). Analysis of variance using the regression coefficient as the test variable demonstrated that the coefficients differed between DenOptix and the remaining systems (p < 0.05) while there were no significant differences between these (p > 0.05). It can be concluded that radiographs obtained with the Dixi and Digora systems were more accurate than Sidexis and DenOptix images for measurement of caries lesion depth. Still, a high probability exists that the individual lesion measurement is either overestimated or underestimated compared to the true lesion depth.