Introduction: This study is aimed to improve the accuracy of cardiac perfusion measurements from sequential ¹³N-Ammonia (¹³NH₃) PET imaging by correcting cardiac and respiratory motion. ¹³NH₃ PET imaging is used to measure cardiac perfusion during the resting and stress state of the heart. Due to the radioactive decay rate of ¹³NH₃, each ¹³NH₃ injection needs to be sufficiently spaced which increases wait times, reduces the quality of care for patients, and increases the cost as imaging time is increased and repeated production of ¹³NH₃ is needed. This can be resolved by sequential ¹³NH₃ injections, followed by carefully subtracting PET images from one injection to another.[1] However, due to cardiac and respiratory motion, these PET images need to be motion corrected before the subtraction. In this study, different cardiac PET motion correction methods are compared for sequential ¹³N-Ammonia PET perfusion measurements.

Methods: A female canine was imaged with a Siemens 3T Biograph mMR PET/MRI scanner, 2 weeks post-acute myocardial infarction. A total of three ¹³NH₃ injections were administered at 0 min, 10 min, and 70 min accompanied by PET listmode and MRI starVIBE acquisition as illustrated in Figure a). During 2^nd and 3^rd 13NH₃ injections, adenosine was also administered to chemically induce cardiac stress. Dynamic PET images were reconstructed with no motion correction, ECG gating, respiratory motion vector correction (BodyCompass^TM) and respiratory motion vector ECG gated motion correction (MV ECG) methods. Further, the 1^st 13NH₃ PET images were decay corrected and subtracted from the second ¹³NH₃ dynamic PET images. The left ventricle (LV) of the heart was segmented and fitted with a two-tissue compartment model to calculate perfusion. [2] Analyses were performed with Carimas, MATLAB and Siemens JSRecon software.[3,4]

Results: Figure b) shows the LV from MRI and four different ¹³NH₃ PET motion correction methods images as mentioned above. First row PET images show standalone stress cardiac ¹³NH₃ PET (3^rd injection). Second row PET images show sequential stressed cardiac ¹³NH₃ PET with subtraction correction (2^nd – 1^st injection). MV ECG shows the thinnest cardiac wall definition compared to other methods. Figure c) shows dynamic relative activity value of LV (both infarcted and non-infarcted region) for standalone and sequential stress cardiac ¹³NH₃ PET.

Discussion: Respiratory and cardiac motion correction by MV ECG enabled the imaging of a specific cardiac cycle which is presented with the thinnest LV thickness definition in PET image. Myocardial perfusion varies throughout the cardiac cycle [5]. Without cardiac motion correction, the LV thickness in PET image is averaged to be thicker as shown in the No motion correction and BodyCompass methods. This cardiac cycle specific perfusion measurement can be translated to the infarcted region of the heart. The perfusion defect of myocardial infarction near the apex of the LV is visible to MRI. The corresponding perfusion defect is clearly demonstrated in the MV ECG method during late diastole phase. Dynamic relative activity of the LV fit to the two-tissue compartment model can quantify myocardial perfusion, however, further optimization is needed when using ECG and MV ECG methods which use a smaller sample of listmode than other methods.

Conclusion: This study demonstrates that cardiac and respiratory motion corrections, using simultaneous PET/MRI, further improve the accuracy of cardiac perfusion measurement when these perfusion measurements are corrected for sequential ¹³NH₃ injections. This will reduce perfusion scan time for cardiac patients. Further modeling is planned to quantify the accuracy of sequential ¹³NH₃ perfusion measurements.

Image/Figure:

Click to view full size

Image/Figure Caption:

Figure. a) sequential rest, stress, stress ¹³N-ammonia PET perfusion experiment protocol. b) comparison between standalone stressed cardiac ¹³NH₃ PET (3^rd injection) and sequential stressed cardiac ¹³NH₃ PET with subtraction correction (2^nd – 1^st injection) for no motion correction, ECG gating, MV ECG and BodyCompass method. c) Dynamic relative activity graph for standalone stressed cardiac ¹³NH₃ PET and sequential stressed cardiac ¹³NH₃ PET with subtraction correction for different motion correction methods.

Author