|
Facebook
Twitter
Linkedin
|
 |
 |
 |
 |
Compression ratios of up to 15:1 offer no risk to diagnostic accuracy
Monday, 29 November 1999 19:00 | 
| Medical Conferences News - RSNA 2007 |
Image compression of CT exams up to 15:1 can significantly reduce storage costs and transfer time without impairing diagnostic accuracy and confidence, a new study has found.
Image compression of CT exams up to 15:1 can significantly reduce storage costs and transfer time without impairing diagnostic accuracy and confidence, a new study has found. The findings were presented this Tuesday at the 2007 Radiological Society of North America (RSNA) annual meeting, in Chicago.
The team from the Medical University of Vienna, Austria, led by Helmut Ringl, MD, retrospectively assessed 104 consecutive emergency abdominal CT (3 mm slice thickness, 2 mm reconstruction interval) examinations of patients with acute abdominal complaints. The images were subjected to lossy, irreversible, three-dimensional JPEG2000 compression using three compression ratios (10:1, 12.5:1 and 15:1).
Three radiologists independently read and reported all patients at all compression ratios. For each organ, readers were asked to state the presence of compression artifacts, the diagnosis and the confidence for diagnosis on five point scale (low=1; high=5). All diagnoses were compared to a gold standard which was built by a abdominal CT expert using the original images, all available surgical reports, follow-ups and patient records.
The team reported that slight to moderate compression artifacts were seen in the liver in 63, 73 and 89 per cent at the compression ratios 10:1, 12.5:1 and 15:1, respectively. The mean percentage of the correct main (=clinical relevant) diagnoses for all three readers were 94.23, 91.56 and 92.57 per cent and did not vary significantly (p=.16) from the original images (94.34 per cent). The mean confidences of all three readers for the applied diagnoses were 4.816, 4.702, 4.745 and did not differ significantly from the original (4.726).
Diagnostic accuracy and confidence was not impaired at compression ratios of up to 15:1, the researchers concluded.
In some countries, the use of lossy compressions is restricted by legal obstacles, but in Austria such restrictions don't exist as no offical assessment has been made yet, Dr Ringl pointed out.
The team from the Medical University of Vienna, Austria, led by Helmut Ringl, MD, retrospectively assessed 104 consecutive emergency abdominal CT (3 mm slice thickness, 2 mm reconstruction interval) examinations of patients with acute abdominal complaints. The images were subjected to lossy, irreversible, three-dimensional JPEG2000 compression using three compression ratios (10:1, 12.5:1 and 15:1).
Three radiologists independently read and reported all patients at all compression ratios. For each organ, readers were asked to state the presence of compression artifacts, the diagnosis and the confidence for diagnosis on five point scale (low=1; high=5). All diagnoses were compared to a gold standard which was built by a abdominal CT expert using the original images, all available surgical reports, follow-ups and patient records.
The team reported that slight to moderate compression artifacts were seen in the liver in 63, 73 and 89 per cent at the compression ratios 10:1, 12.5:1 and 15:1, respectively. The mean percentage of the correct main (=clinical relevant) diagnoses for all three readers were 94.23, 91.56 and 92.57 per cent and did not vary significantly (p=.16) from the original images (94.34 per cent). The mean confidences of all three readers for the applied diagnoses were 4.816, 4.702, 4.745 and did not differ significantly from the original (4.726).
Diagnostic accuracy and confidence was not impaired at compression ratios of up to 15:1, the researchers concluded.
In some countries, the use of lossy compressions is restricted by legal obstacles, but in Austria such restrictions don't exist as no offical assessment has been made yet, Dr Ringl pointed out.
