Evaluation of the quantitative accuracy of a commercially available positron emission mammography scanner

The PEM Flex Solo II (Naviscan, Inc., San Diego, CA) is currently the only commercially available positron emission mammography (PEM) scanner. This scanner does not apply corrections for count rate effects, attenuation, or scatter during image reconstruction, potentially affecting the quantitative accuracy of images. The aim of this work is to measure the overall quantitative accuracy of the PEM Flex system and to determine the individual contributions of error from count rate effects, attenuation, and scatter.

Gelatin phantoms were designed to simulate breasts of different thicknesses (4-12 cm) with varying uniform background activity concentration (AC) (0.007-0.5 microCi/cc), cysts, and lesions (2:1, 5:1, and 10:1 lesion-to-background ratios). The overall error was calculated from ROI measurements in the phantoms with a clinically relevant background AC (0.065 microCi/cc). The error due to count rate effects was determined by comparing the overall error at multiple background AC to the error at a very low background AC (0.007 microCi/cc) where count rate effects are considered negligible. A point source and cold gelatin phantoms of different thicknesses were used to assess the errors due to attenuation and scatter. The maximum pixel values of the point source in gelatin and in air were compared to determine the effect of attenuation, while scatter was evaluated by comparing the sum of all pixel values in gelatin and in air.

The AC in the uniform background was underestimated in phantoms of all thicknesses, with the exception of the 4-cm-thick phantoms in which the measured AC was within +/- 7% of the true value (0.065 microCi/cc). The degree of underestimation in the uniform background increased with the phantom thickness, up to 34% +/- 6% in the 12 cm phantoms. The AC in all lesions was underestimated by more than that measured in the background (22% for the 2:1 lesions in the 4 cm phantom) and this underestimation increased with increasing thickness and lesion-to-background ratio (85% for the 10:1 lesions in the 12 cm phantoms). The error due to count rate effects reduced the measured AC with respect to its true value. This error increased with increasing background AC (23% +/- 6% at 0.065 microCi/cc to 85% +/- 1% at 0.5 microCi/cc for the 4 cm phantoms) and decreased with increasing phantom thickness (23% +/- 6% for the 4-cm-thick phantoms to 7% +/- 7% for the 12-cm-thick phantoms at 0.065 microCi/cc). Attenuation was a substantial source of error and reduced the measured AC by 51% +/- 10%-77% +/- 4% in the 4-12 cm phantoms, respectively. Scatter increased the total signal measured in images by a relatively constant amount (23% +/- 9%) for all thicknesses.

Applying corrections for count rate effects, attenuation, and scatter will be essential for the PEM Flex Solo II to be able to produce quantitatively accurate images.