OBJECTIVE To evaluate the accuracy of proximal caries detection comparing enhanced and unenhanced Siemens Sidexis CCD-based digital images with Ektaspeed Plus films utilizing receiver operating characteristic analysis. STUDY DESIGN Sixty extracted teeth (24 posterior and 36 anterior) were imaged under identical standardized geometric and exposure conditions. Six observers, using a 5-point confidence scale, rated 120 proximal surfaces for the presence or absence of carious lesions by means of three image modalities: (1) observer enhanced and (2) unenhanced Sidexis displays, and (3) Ektaspeed Plus films. The ground truth was determined by microscopic analysis of ground sections. Receiver operating characteristic curves were generated with calculated areas (AZ) analyzed with analysis of variance for effect of reader, reading, and modality. RESULTS Analysis of variance demonstrated significant differences among readers, readings and modalities (mean square values of 0.012, 0.005, 0.004, F ratios of 13.604, 5.329, 5.100; p = values of 0.001, 0.043, and 0.030, respectively). Post-hoc paired comparisons of modalities using Tukey's statistic demonstrated that only film and enhanced Sidexis images were different from each other (p = 0.024). AZ scores were 0.7650, 0.7499, and 0.8008 respectively, for unenhanced Sidexis, enhanced Sidexis, and Ektaspeed Plus film. CONCLUSION Unenhanced digital Sidexis images were equivalent to film for the detection of proximal caries in this in vitro study. Observer enhanced Sidexis images exhibited a statistically significant lower diagnostic accuracy than the unenhanced digital and film images.