Activity quantification of 177Lu using a 360◦ CZT gamma camera - comparison with a dual-head Anger camera

Sammanfattning: Background and Aim: New SPECT cameras that use CZT detectors in a 360◦ configuration has recently been introduced. Together with Bayesian reconstruction methods employing Median Root Prior (MRP) and Relative Difference Prior (RDP), this constitutes an alternative to Anger-based SPECT for activity quantification. The aim of this thesis is to, by the use of both OS-EM and Bayesian reconstruction methods, assess the feasibility of 360◦ CZT SPECT (GE Starguide™) for activity quantification of 177Lu, using activity quantification of 99mTc as a baseline and compare its performance to that of an Anger-based SPECT camera (GE Discovery NM/CT 670, 5/8" crystal). Method: To compare the performance of the cameras for quantitative SPECT, various phantoms in different geometries were imaged in both cameras. First, cylindrical uniform phantoms with 177Lu or 99mTc were imaged. The images of the uniform phantom with 177Lu was used to optimise the Signal-to-Noise ratios (SNR) of the 177Lu energy and scatter windows for 360◦ CZT SPECT and compare these to Anger-based SPECT. The dependency of the reconstructed signal on position in the Field-Of-View (FOV) of the 360◦ CZT SPECT was investigated by testing the difference in means of reconstructed signal between Volume-Of-Interests (VOIs) placed in different transaxial and axial positions of the uniform phantoms, for both radionuclides. The dependency of the reconstructed signal on geometry was subsequently investigated by repeated imaging of the same spherical source in different parts of FOV. Then the difference in means of reconstructed signal between the delineated spheres in the different images was tested. Additionally, a point source of 177Lu or 99mTc was imaged repeatedly by using different source-to-collimator distances. Finally, calibration factors for 99mTc and both peaks of 177Lu was determined for both systems by relating reconstructed signal to known phantom activity. These calibration factors were then used to compute recovery for spheres, as function of volume in a NEMA body phantom. Results: The results show that the reconstructed signal of the 360◦ CZT SPECT has a positional dependency for 99mTc and both peaks of 177Lu in transaxial and axial directions. The optimal energy window of the 360◦ CZT SPECT for 177Lu, was asymmetric and was centred at 205.3 keV with a ±5.5 % width using scatter compensation with Triple-Energy-Window(TEW). This energy session was found to achieve improved SNR compared to the Anger-based SPECT for the same acquisition time and activity. The calibration factor of this energy session, regardless of reconstruction method, was found to be 80.2 s−1 MBq−1 but had a significant positional dependency (p = 1 × 10−4 ) with a 18 % difference at most between the calibration factors found when placing the VOI used for calibration factor determination in different transaxial positions of the uniform phantom. The same energy session used for imaging of a sphere at different FOV positions was similarly found to result in a significant variation in reconstructed signal (p = 2 × 10−17) with a 9 % difference at most. The point source acquisitions with a varying source-to-collimator distance showed that there was a slight difference in reconstructed signal with different FOV sizes. The recovery of 177Lu and 99mTc in spheres with different volumes for Anger-based SPECT followed expected trends while the recovery for 360◦ CZT SPECT displayed unreasonably high recovery for spheres positioned in the upper part of FOV. Conclusion: The reconstructed signal of 360◦ CZT SPECT does display a dependency on geometry which makes its use for activity quantification inaccurate. The same dependency is not seen for Anger-based SPECT, which displayed stable calibration factors thus allowing activity quantification with increased accuracy. The 360◦ CZT SPECT does however, compared to Anger-based SPECT, exhibit an increased SNR as well as a higher recovery, although unstable. Therefore, it displays promising results, should the reason for the unstability be discovered and solved.