Abstracts

Rationale: SPECT imaging is a type of nuclear medicine exam which uses a radiotracer (Tc-99m) to track regional cerebral brain perfusion. The combination of Ictal and Interictal SPECT imaging can be useful in identifying the epileptogenic zone during the course of a presurgical evaluation. By comparing a baseline interictal scan to an ictal scan, the areas of seizure induced hyper-perfusion can be identified and coregistered to high resolution MRI data (SISCOM). Our institution recently introduced this imaging service into our EMU, however our initial results were not matching the yields reported in the literature. Much of the literature stresses the importance of a quick (less than 20s post seizure) injection, as there is a compensatory rebound effect that will result in relative hypo-perfusion if the injection time is too late. All fifteen of the initial patients received injections in less than 20s due to the use of a power injector system and video EEG monitoring. However, only 5/15 of the initial patients had useful SISCOM results, which falls well below the reported yield of 60-80% for this diagnostic exam. Upon further review, we were able to identify that the problem was not due to the injection time, patient selection, or amount of radiotracer administered. Although we were using an FDA approved 3rd party commercial software solution for our image processing, it became evident that we needed to systematically validate the image acquisition and post-processing techniques. Methods: A nuclear medicine phantom was created using glass pipettes filled with Tc-99m, and assembled using known geometries. By using pipettes of different lengths and spacing them in orthogonal directions, we were able to measure the geometric accuracy of our imaging chain in all three dimensions. We then scanned the phantom using both GE Discovery and Siemens Symbia scanners, as well as switching between low energy high resolution (LEHR) and fan beam collimators. Each of the scans were reconstructed directly on the scanner as well with the 3rd party software. Measurements were then made using ImageJ, and analyzed for geometric accuracy. Results: Using the LEHR collimators, there was very little geometric error in the reconstructions. The Siemens LEHR scans resulted in an error of 4% in the Z direction and 2% in the X-Y plane for both the scanner produced reconstructions and the 3rd party reconstructions. However, when using the fan beam collimators the results were not consistent. The individual vendors were able to accurately reconstruct their own data on the scanner stations: Siemens fan beam ( 4% in Z, 1% in X-Y), and GE fan beam (0% in both directions). However, the 3rd party software was not able to reconstruct the data appropriately: 3rd party reconstructed Siemens fan beam data (3% in Z, 30% in X-Y), and 3rd party reconstructed GE fan beam data (7% in Z, and up to 60% in X-Y depending on recon method). Conclusions: These results provide justification for acceptance testing of all software processing techniques used in the clinic regardless of vendor or FDA approval. Despite being FDA approved, a small error in the 3rd party software led to major differences in the reconstruction of the SPECT data. Upon recognizing these errors, we used alternative methods which were thoroughly validated to reprocess the data. After reprocessing the original data, 13/15 of the initial patients showed useful SISCOM results. It should also be noted that the 3rd party software vendor corrected their error upon notification. Funding: none