A technique for the simulation of planar radionuclide images of the kidney

Although Tc-dimercaptosuccinic acid (DMSA) scans are routinely used to quantify relative renal function, no quantification method is universally adopted. Audits using real patient data indicate reasonable consistency but, as the true relative function is unknown, accuracy cannot be assessed. The aim was to simulate realistic DMSA images that can be used to assess accuracy.

Anatomical models were created from computed tomography (CT) scans of a patient who had also undergone DMSA imaging. Organs that take up DMSA were outlined on CT and each assigned an activity concentration (with renal cortex and medulla modelled separately). The simulated images were visually compared to the patient's clinical images and subtracted to identify differences. Iteration was used on the posterior image to find the organ activities that produced the most realistic simulated image. The optimal activity distribution was then used to also simulate an anterior image. To assess the simulations, the percentage difference was calculated between the counts in each kidney on the real and simulated images.

Visually, the clinical and simulated images appear similar and the subtracted images indicate only small differences. The percentage difference in kidney counts between the images was less than 1% for both kidneys on the posterior image and less than 5% on the anterior image. The cortex and medulla activity concentrations were approximately equal.

A technique for realistic simulation of DMSA images has been devised and should prove useful for evaluating image analysis software.