Methods: Periapical images of two phantoms (mandible/maxilla) were taken at three kilovoltages (60, 70 and 80kVp) with mAs adjusted to produce isodensity. Images were digitized (400dpi, 8-bit) and stored as: 1) TIF format uncompressed; 2) JPEG (1992) [J92] and; 3) JPEG2000 [J2K]. Images were compressed at four levels (69.4%, 93%, 95.7% and 97.1% file size reduction). Five general dentists viewed film and digital images, including nine re-reads. Subjective image assessment (SIA) based on a 5-point Likert scale was determined from averaging observations of the image quality of six regions of interest. Entire image and localized dentin signal-to-noise ratios (SNRe and SNRd) as well as the IQM (Image Quality Measure, v6.2) of each image was calculated. The effect of kVp, type and level of compression on SIA and IQM was plotted and compared. Correlation coefficients (r) and regression equations were determined and compared at a significance level of p£0.05.
Results: Intra-observer reliability was very good (r=.89). SIA reduced below acceptable levels with increasing compression level beyond 93% for the mandibular image and beyond 95.7% for the maxillary image. Maxillary SIA was significantly greater at all levels of compression except 95.7%. The IQM range for the maxilla was less than that for the mandible. Overall IQM provided relatively consistent regression plots and demonstrated significantly higher correlation as a Log10 function with SIA than SNRe or SNRd. Similarly IQM provided significantly better correlation than either SNRe or SNRd for the maxilla, J92 and J2K compressions.
Conclusions: These results suggest that IQM values vary depending on the image analyzed. IQM shows promise as a quantitative index of image quality with very good correlation to SIA, superior to either SNRe or SNRd.