The effects of image compression on quantitative measurements of digital panoramic radiographs

Abstract

Objectives: The aims of this study were to explore how image compression affects density, fractal dimension, linear and angular measurements on digital panoramic images and assess inter and intra-observer repeatability of these measurements.

Study design: Sixty-one digital panoramic images in TIFF format (Tagged Image File Format) were compressed to JPEG (Joint Photographic Experts Group) images. Two observers measured gonial angle, antegonial angle, mandibular cortical width, coronal pulp width of maxillary and mandibular first molar, tooth length of maxillary and mandibular first molar on the left side of these images twice. Fractal dimension of the selected regions of interests were calculated and the density of each panoramic radiograph as a whole were also measured on TIFF and JPEG compressed images. Intra-observer and inter-observer consistency was evaluated with Cronbach's alpha. Paired samples t-test and Kolmogorov-Smirnov test was used to evaluate the difference between the measurements of TIFF and JPEG compressed images.

Results: The repeatability of angular measurements had the highest Cronbach's alpha value (0.997). There was statistically significant difference for both of the observers in mandibular cortical width (MCW) measurements (1st ob. p: 0.002; 2nd ob. p: 0.003), density (p<0.001) and fractal dimension (p<0.001) between TIFF and JPEG images. There was statistically significant difference for the first observer in antegonial angle (1st ob p< 0.001) and maxillary molar coronal pulp width (1st ob. p< 0.001) between JPEG and TIFF files.

Conclusions: The repeatability of angular measurements is better than linear measurements. Mandibular cortical width, fractal dimension and density are affected from compression. Observer dependent factors might also cause statistically significant differences between the measurements in TIFF and JPEG images.

Figure 1

1) The original region of interest which was created for fractal dimension calculations. 2) Blurred form of the region of interest by applying a Gaussian filter with a diameter of 20. 3) Binary form of the region of interest with which fractal dimension was calculated.

Figure 2

Cropped panoramic radiograph showing the gonial angle (1), antegonial angle (2) and mandibular cortical width (3).

Figure 3

Cropped and rotated digital panoramic radiograph showing the measurement of maxillary (2) and mandibular (4) coronal pulp width, maxillary (1) and mandibular (3) tooth length and the region of interest placement (5) for fractal dimension calculations.