Improvement in DMSA imaging using adaptive noise reduction: an ROC analysis

Dimercaptosuccinic acid imaging is the 'gold standard' for the detection of cortical defects and diagnosis of scarring of the kidneys. The Siemens planar processing package, which implements adaptive noise reduction using the Pixon algorithm, is designed to allow a reduction in image noise, enabling improved image quality and reduced acquisition time/injected activity. This study aimed to establish the level of improvement in image quality achievable using this algorithm. Images were acquired of a phantom simulating a single kidney with a range of defects of varying sizes, positions and contrasts. These images were processed using the Pixon processing software and shown to 12 observers (six experienced and six novices) who were asked to rate the images on a six-point scale depending on their confidence that a defect was present. The data were analysed using a receiver operating characteristic approach. Results showed that processed images significantly improved the performance of the experienced observers in terms of their sensitivity and specificity. Although novice observers showed significant increase in sensitivity when using the software, a significant decrease in specificity was also seen. This study concludes that the Pixon software can be used to improve the assessment of cortical defects in dimercaptosuccinic acid imaging by suitably trained observers.