Investigating the limit of detectability of a positron emission mammography device: A phantom study

Purpose: A new positron emission mammography (PEM) device (PEM Flex Solo II, Naviscan Inc., San Diego, CA) has recently been introduced and its performance characteristics have been documented. However, no systematic assessment of its limit of detectability has been evaluated. The aim of this work is to investigate the limit of detectability of this new PEM system using a novel, customized breast phantom. Methods: Two sets of F-18 infused gelatin breast phantoms of varying thicknesses (2, 4, 6, and 8 cm) were constructed with and without (blank) small, shell-less contrast objects (2 mm thick disks) of varying diameters (3-14.5 mm) volumes: 0.15-3.3 cc and activity concentration to background ratio (ACR) (2.7-58). For the phantom set with contrast objects, the disks were placed centrally inside the phantoms and both phantom sets were imaged for a period of 10 min on the PEM device. In addition, scans for the 2 and 6 cm phantoms were repeated at different times (0, 90, and 150 min) post phantom construction to evaluate the impact of total activity concentration (count density) on lesion detectability. Each object from each phantom scan was then segmented and placed randomly in a corresponding blank phantom image. The resulting individual images were presented blindly to seven physician observers (two nuclear medicine and five breast imaging radiologists) and scored in a binary fashion (1-correctly identified object, 0-incorrect). The sensitivity, specificity, and accuracy of lesion detectability were calculated and plots of sensitivity versus ACR and lesion diameters for different phantom thicknesses and count density were generated. Results: The overall (mean) detection sensitivity across all variables was 0.68 (95% CI: 0.64, 0.72) with a corresponding specificity of 0.93 0.87, 0.98, and diagnostic accuracy of 0.72 0.70, 0.75. The smallest detectable object varied strongly as a function of ACR, as sensitivity ranged from 0.36 0.29, 0.44 for the smallest lesion size (3 mm) to 0.80 0.75, 0.84 for the largest (14.5 mm). Conclusions: The detectability performance of this PEM system demonstrated its ability to resolve small objects with low activity concentration ratios which may assist in the identification of early stage breast cancer. The results of this investigation can be used to correlate lesion detectability with tumor size, ACR, count rate, and breast thickness.