John Anthony Kennedy MS, Abstract Co-Author: Nothing to Disclose

Ora Israel MD, Abstract Co-Author: Nothing to Disclose

Alex Frenkel DSC, Abstract Co-Author: Nothing to Disclose

Rachel Bar-Shalom MD, Abstract Co-Author: Nothing to Disclose

Haim Azhari DSc, Presenter: Nothing to Disclose

In clinical PET scanners the intrinsic spatial resolution along the axial (bed) direction is often lower than the resolution in the transaxial plane. The purpose of the present study was to evaluate and implement a “super-resolution” method for improving the resolution in the axial direction with no changes in hardware.

“Super-resolution” was obtained by combining four data sets acquired with Ľ pixel axial shifts between consecutive acquisitions and applying an iterative algorithm to generate high-resolution images. The technique was evaluated in both phantom and patient studies. Using a clinical PET scanner (Discovery LS, GEMS), a “super-resolution” scan of a special phantom containing sets of sources ranging from 1 to 8 mm in diameter was obtained. In a patient study, a region of interest containing a small lesion was chosen following a standard full-body 18F-FDG PET/CT scan. Four additional PET acquisitions shifted axially by Ľ pixel between acquisitions of the field of view (FOV) containing the lesion were acquired. This was performed without any further increase in patient radiation exposure. The “super-resolution” algorithm was applied to generate images of this FOV with Ľ the original slice thickness. In addition, to check consistency, the “super-resolution” images were degraded to allow comparison with the original low-resolution images.

In phantom studies, application of the algorithm demonstrated that while only 6 mm features of the phantom could be resolved in the original images, the proposed “super-resolution” technique clearly resolved 3 mm features. In the patient study, details less than 4 mm in the small lesion were observed in the “super-resolution” reconstructions which were not observed in the original images of the lesion. Degradation of the “super-resolution” results yielded images comparable to the original, indicating that the algorithm is consistent with a standard PET acquisition.

A new approach for improving the axial resolution of PET images using a “super-resolution” method has been developed and experimentally confirmed. The method can be implemented on clinical scanners without changing the hardware.

Kennedy, J, Israel, O, Frenkel, A, Bar-Shalom, R, Azhari, H, “Super-Resolution” in PET Imaging.  Radiological Society of North America 2004 Scientific Assembly and Annual Meeting, November 28 - December 3, 2004 ,Chicago IL. http://archive.rsna.org/2004/4408882.html