Unencrypted login | home

Program Information

Using Quantitative Measurement of Image Quality to Standardize Clinical Performance Across a Diverse Population of CT Scanners

J Och

J Och*, M.S., W Kresge, M.S., D Snyder, M.S., J. Lock, M.D., S. Moshiri, M.D.,Geisinger Medical Center, Danville, PA

SU-E-I-54 Sunday 3:00PM - 6:00PM Room: Exhibit Hall

Purpose: To develop an Image Quality metric, in order to establish consistent Image Quality among a population of CT scanners, of different make/model and locations in a health system.

Methods:Through discussion with radiologists, it was decided to use image noise as a metric. Image noise was defined as the standard deviation of an ROI placed in the image. Radiologists selected target values for the desired level of image noise.
For head examinations, the ROI was placed in the solid tissue at the level of the fornix. For abdomen images, the ROI was placed at mid-liver. For the actual metric, the average of two ROI's was used.

For each scanner in the system, two data sets (brain and abdomen) of 100 images each were evaluated. Population statistics were acquired and the range of variation was measured. Data analysis was done on the entire population of images, as well as subsets, specifically location and scanner make/model.

Based on the results of the analysis, protocols were revised, where necessary, to address excessive variation, and to correlate with the desired image noise target, 4.0 for the brain and 11.5 for the abdomen.

Results:Use of the application resulted in a decrease in the range of image noise for brain images from 3.6 - 5.5 to 4.1 - 4.9. The range for abdomen images decreased from 7.0 - 17.4 to 9.6 - 14.1.
Average noise values for abdomen images decreased from 12.3 to 11.8. Average Image Noise for the brain decreased from 4.6 to 4.3.
The initial difference in average image noise between the two main scanner populations was 2.9 for abdomen scans. After the project the interscanner abdomen image noise difference decreased to 1.3.

Conclusion: Image noise is a viable metric for comparing scanner performance and standardizing protocols.

Contact Email: