^{1,a)}, William D. Erwin

^{1}and Richard E. Wendt III

^{1}

### Abstract

An alternative to the conventional method of performing the AAPM Report 52 rotational uniformity and sensitivity test has been developed. In contrast to the conventional method in which a Co-57 sheet source is fastened to the collimator, this new point-source method acquires the images intrinsically using a Tc-99m point source placed near the isocenter of gantry rotation. As with the conventional method, the point-source method acquires count flood images at four distinct gantry positions to calculate the maximum sensitivity variation (MSV)—a quantitative metric of rotational uniformity and sensitivity variation. The point-source method incorporates corrections for the decay of Tc-99m between acquisitions, the curvature in the image intensity due to variation in photon flux across the detector from a near-field source, and the source-to-detector distance variations between views. The raw point-source images were fitted with an analytic function in order to compute curvature- and distance-corrected images prior to analysis. Five independent MSV measurements were performed using both conventional and point-source methods on a single detector of a dual-headed SPECTsystem to estimate the precision of each method. The precision of the point-source method was further investigated by performing ten independent measurements of MSV on six different detectors. Correlation between the MSV calculated by the two methods was investigated by performing the test on nine different detectors using both methods. Different levels of sensitivity variations were also simulated on four detectors to generate 40 additional paired points for correlation analysis. The effect of the total image counts on the MSV estimated with the new method was evaluated by acquiring image sequences with , , and count images. The MSV calculated using the conventional and point-source methods exhibited a high degree of correlation and consistency with equivalence. The precision of the point-source method (0.145%) is lower than the conventional method (0.04%) but sufficient to test MSV. No statistically significant dependence of MSV with the point-source method on the total image counts over a range of counts was observed. Curvature correction of the images prior to the generation of difference images renders images more conducive to qualitative inspection for structured, nonrandom patterns. The advantages of the new methodology are that multiple detectors of a gamma camera can be evaluated simultaneously which substantially reduces the time required for MSV testing and the reduced risk of accidental damage to the collimators and patient proximity detectionsystem from having to mount a sheet source on each of the detectors.

I. INTRODUCTION

II. MATERIALS AND METHODS

II.A. Accuracy of source-to-detector distance estimate from the point-source image model

II.B. Precision of the MSV calculations

II.C. Correlation between the MSV calculated with the two methods

II.D. Dependence of MSV calculation on total image counts for the point-source method

III. RESULTS

III.A. Accuracy of source-to-detector distance estimate from the point-source image model

III.B. Precision of the MSV calculations

III.C. Correlation between the MSV calculated with the two methods

III.D. Dependence of MSV calculation on total image counts for the point-source method

IV. DISCUSSION

IV.A. Accuracy of source-to-detector distance estimate from the point-source image model

IV.B. Precision of the MSV calculations

IV.C. Correlation between the MSV calculated with the two methods

IV.D. Qualitative assessment of curvature-corrected images

IV.E. Advantages of the point-source method compared to the conventional method

V. CONCLUSION

### Key Topics

- Image sensors
- 99.0
- Medical imaging
- 29.0
- Single photon emission computed tomography
- 18.0
- Image detection systems
- 11.0
- Cameras
- 10.0

## Figures

The source-to-detector distances estimated from the curvature-correction fit to the point-source image plotted against the true distances at four different distances for two different detectors.

The source-to-detector distances estimated from the curvature-correction fit to the point-source image plotted against the true distances at four different distances for two different detectors.

The maximum sensitivity variation calculated with the point-source method plotted against that for the conventional method. The error bars shown are the precision estimates for each of the two methods (Sec. III C and Table I). (a) Paired points for the measured data from nine different detectors. The dashed line plotted is the line of equality with slope of unity and intercept of zero. (b) Paired points for the measured data combined with simulated data. The dashed line plotted is the fit of the data to a straight line.

The maximum sensitivity variation calculated with the point-source method plotted against that for the conventional method. The error bars shown are the precision estimates for each of the two methods (Sec. III C and Table I). (a) Paired points for the measured data from nine different detectors. The dashed line plotted is the line of equality with slope of unity and intercept of zero. (b) Paired points for the measured data combined with simulated data. The dashed line plotted is the fit of the data to a straight line.

A Bland-Altman plot of agreement in MSV calculated with both point-source and conventional methods. The graph plots the mean of the two measurements as the abscissa and the difference between them (point-source minus conventional) as the ordinate. The data plotted include MSV for the nine detectors (identified by square boxes) and the simulated data. The mean difference (bias) in the point-source method and the SD limits are also indicated.

A Bland-Altman plot of agreement in MSV calculated with both point-source and conventional methods. The graph plots the mean of the two measurements as the abscissa and the difference between them (point-source minus conventional) as the ordinate. The data plotted include MSV for the nine detectors (identified by square boxes) and the simulated data. The mean difference (bias) in the point-source method and the SD limits are also indicated.

A demonstration of the curvature-correction algorithm for a near-field point-source image as developed in the Appendix: (a) The uncorrected raw image, (b) the corresponding curvature-corrected image generated after fitting the image to the point-source image model, and (c) the center profiles through the raw and the curvature-corrected images.

A demonstration of the curvature-correction algorithm for a near-field point-source image as developed in the Appendix: (a) The uncorrected raw image, (b) the corresponding curvature-corrected image generated after fitting the image to the point-source image model, and (c) the center profiles through the raw and the curvature-corrected images.

A demonstration of the use of curvature-corrected images for the qualitative inspection of structured, nonrandom patterns in subtracted images from two detectors. Note that neither detector has any real artifacts that should be seen. The differences (0° image subtracted from the 90°, 180°, 270°, and 360° images) for the point-source method are shown: [(a) and (c)] Without curvature correction prior to subtraction and [(b) and (d)] with curvature correction prior to subtraction.

A demonstration of the use of curvature-corrected images for the qualitative inspection of structured, nonrandom patterns in subtracted images from two detectors. Note that neither detector has any real artifacts that should be seen. The differences (0° image subtracted from the 90°, 180°, 270°, and 360° images) for the point-source method are shown: [(a) and (c)] Without curvature correction prior to subtraction and [(b) and (d)] with curvature correction prior to subtraction.

A schematic of the point-source imaging geometry for the proposed point-source methodology for testing rotational uniformity and sensitivity variations. Distances and variables shown are discussed at depth in the Appendix. The distances are not shown to scale but adjusted for clarity (e.g., ).

A schematic of the point-source imaging geometry for the proposed point-source methodology for testing rotational uniformity and sensitivity variations. Distances and variables shown are discussed at depth in the Appendix. The distances are not shown to scale but adjusted for clarity (e.g., ).

## Tables

The five separate and independent measurements of the maximum sensitivity variation (MSV%) for the same detector using both point-source and the conventional methods together with their mean values, SD, and coefficient of variation (COV).

The five separate and independent measurements of the maximum sensitivity variation (MSV%) for the same detector using both point-source and the conventional methods together with their mean values, SD, and coefficient of variation (COV).

The ten measurements of the maximum sensitivity variation (MSV%) for six different detectors [A–D: 9.5 mm crystal; E–F: 15.9 mm crystals] using the point-source method together with their mean values, SD, COV, variance, minimum, and maximum values.

The ten measurements of the maximum sensitivity variation (MSV%) for six different detectors [A–D: 9.5 mm crystal; E–F: 15.9 mm crystals] using the point-source method together with their mean values, SD, COV, variance, minimum, and maximum values.

The maximum sensitivity variation (MSV%) for two different detectors (A and B) with the point-source method using image sequences with , , and counts per image. The SD and range of MSV calculated are also shown.

The maximum sensitivity variation (MSV%) for two different detectors (A and B) with the point-source method using image sequences with , , and counts per image. The SD and range of MSV calculated are also shown.

Article metrics loading...

Full text loading...

Commenting has been disabled for this content