Per-image encoding (a) and decoding (b) times of 2D and 3D compressions. Light and dark gray bars represent 2D and 3D compressions, respectively. Error bars indicate 95% confidence intervals.
PSNR values at each compression ratio for 2D and 3D compressions. Light and dark gray bars represent PSNR for 2D and 3D compressions, respectively. Error bars indicate 95% confidence intervals.
JPEG2000 2D and 3D compression artifacts in a contrast-enhanced transverse abdomen CT image (top) in a 67-year-old female with acute cholecystitis. The subtraction images for 2D compression (left column) and for 3D compression (right column) represent the mathematical differences between the original and compressed images at each irreversible compression ratio (4:1, 6:1, 8:1, 10:1, and 12:1, from top to bottom). Note that a subtraction map for the 3D compression at a given tested compression ratio is similar to or brighter (indicating more artifacts) than that for the 2D compression at the adjacent lower ratio. The compression artifacts are best demonstrated if the original and compressed images (supplementary materials) are displayed alternately on a calibrated monitor.
Grading responses of reader 1 (a), reader 2 (b), and reader 3 (c). Two bars at each compression ratio represent 2D (left) and 3D (right) compressions, respectively. Each gray shade indicates a different grade for the artifacts (white, grade 0; light gray, grade 1; dark gray, grade 2; and black, grade 3).
Visually lossless thresholds (VLTs) determined by pooled radiologists’ binary responses. Solid line bubbles indicate that the VLTs of 3D compression matched VLTs of 2D compression. Dashed line bubbles indicate that VLTs of 3D compression were higher than those of 2D compression. The bubble area is proportional to the number of superimposed data points.
Pathologies contained in the 60 CT scans forming the study sample.
Readers’ grading responses. [Note: data are of the artifact grading for the 60 tested images. A -value (adjusted by Bonferroni correction) indicates a significant difference. A dash (-) indicates that a -value could not be calculated.]
Readers’ binary responses. [Note: data are the number of image pairs (and percentage of the 60 image pairs) that were distinguishable. A -value (adjusted by Bonferroni correction) indicates a significant difference.]
Article metrics loading...
Full text loading...