ISSN:
1619-7089
Keywords:
Gamma-camera
;
Quality control
;
Uniformity
;
Cusum analysis
Source:
Springer Online Journal Archives 1860-2000
Topics:
Medicine
Notes:
Abstract Gamma-camera uniformity is often monitored for quality control of performance. Many parameters can be derived from computer analysis of the image of a uniform (flood) source to obtain quantitative measures of uniformity. However, it is not clear which best reflect changes in uniformity, or how action levels should be set. The latter are essential for an automated analysis. To compare the various parameters and assess two techniques used for their analysis, a series of technetium-99m flood images were obtained consecutively. After establishing a baseline from repeated images, uniformity was degraded by offsetting the pulse height analyser (PHA) window by a small amount. Nine non-uniformity parameters were calculated for each flood image, including the Department of Health and Social Security (DHSS) and National Electrical Manufacturers Association (NEMA) integral and differential uniformities, the uniformity index and the percentage of pixels more than 5% from the mean. All parameters were calculated for both central field-of-view (CFOV) and useful field-of-view (UFOV). Eighty such trials were carried out on two gamma-cameras. The PHA offsets ranged from 1–10 keV The two techniques used for analysis were the control chart and cusum analysis. Decision levels were set for each parameter, as multiples of SD for the control charts or by varying the dimensions of a truncated V-mask for cusum analysis. Receiver operating characteristic (ROC) curves were then used for an objective comparison of the parameters and of the analysis techniques. The area under the ROC curve was used to rank the parameters in order of sensitivity to change in uniformity. The uniformity index and NEMA integral non-uniformity were found to be the most sensitive. Cusum analysis improved the sensitivity of those parameters that appeared to have low sensitivity to change when subjected to control chart analysis. However, there was little difference between the two analysis techniques for the more sensitive parameters. The ROC curves could also be used to provide a method for selecting decision levels, hence establishing an automated method for gamma-camera quality control.
Type of Medium:
Electronic Resource
URL:
http://dx.doi.org/10.1007/BF00184128
Permalink
