Effect of Digital Contrast Optimization on the Detection of Caries

Methods: A panel of three experienced radiologists selected 100 correctly exposed bitewings. The radiographs were scanned on a flatbed scanner and the average cumulative grey-level histogram (ACGH) was calculated for these radiographs. Also, 50 over- and 50 underexposed films were selected and scanned. An expert panel viewed the original radiographs to establish ground truth about the presence and depth of caries lesions (because the radiographs had been taken from clinical patients, no histology could be done). The grey level distribution of each of the over- or underexposed images was optimised using the ACGH of the correctly exposed images as a reference (method 1) and using linear contrast stretching, excluding grey levels representing metal restorations (extreme radiopaque) and background (extreme radiolucent) (method 2). Eight observers assessed the images in random order; results were scored as observed lesion depth and observer confidence. Statistical differences were analysed using General Linear Models (SPSS 9.0) and considered significant when p < 0.05.

Results: Linear contrast stretching performed better than the cumulative histogram optimisation (p=0.04). Both were better than the original (p=0.02). The benefit of contrast optimisation was larger for underexposed radiographs than for overexposed radiographs.

Conclusions: Contrast optimisation in general improves the diagnostic image quality in caries detection. Simple linear contrast stretching performed better than the cumulative histogram optimisation.