OBJECTIVES: To determine the effect of noise on the compressibility and the diagnostic accuracy for caries detection of digital bitewing radiographs. METHODS: Bitewing radiographs of patients were obtained with a storage phosphor (Digora, Soredex, Helsinki, Finland) and compressed at different JPEG compression levels (2, 27, 53 and 128). A just noticeable difference study was performed to select a compression level to study the added noise effect. Gaussian noise was added at low, medium, and high levels to both the original and compressed images. Seven observers examined the selected approximal surfaces to identify the caries depth. ROC analysis was performed together with ANOVA at P = 0.05. RESULTS: The compressibility of the images decreased as the noise level increased. ROC analysis revealed no significant difference between the original and compressed images within the same noise level (P > 0.06). With added noise compressed/decompressed images had a higher Az than the corresponding original images. CONCLUSIONS: JPEG compression at level 27 can be used without a significant deterioration in diagnostic accuracy. Compression at this level seemed to reduce the effect of noise to some extent.
OBJECTIVES: To determine the effect of noise on the compressibility and the diagnostic accuracy for caries detection of digital bitewing radiographs. METHODS: Bitewing radiographs of patients were obtained with a storage phosphor (Digora, Soredex, Helsinki, Finland) and compressed at different JPEG compression levels (2, 27, 53 and 128). A just noticeable difference study was performed to select a compression level to study the added noise effect. Gaussian noise was added at low, medium, and high levels to both the original and compressed images. Seven observers examined the selected approximal surfaces to identify the caries depth. ROC analysis was performed together with ANOVA at P = 0.05. RESULTS: The compressibility of the images decreased as the noise level increased. ROC analysis revealed no significant difference between the original and compressed images within the same noise level (P > 0.06). With added noise compressed/decompressed images had a higher Az than the corresponding original images. CONCLUSIONS: JPEG compression at level 27 can be used without a significant deterioration in diagnostic accuracy. Compression at this level seemed to reduce the effect of noise to some extent.