Improved Partial Volume Correction for PET Myocardial Blood Flow Imaging

Introduction: Partial volume correction (PVC) for PET myocardial blood flow (MBF) quantification is typically accomplished by the inclusion of a recovery factor (1 − v_b) in the one-tissue kinetic model: C(t) ≈ (1 − v_b )K₁e^-k₂t ∫₀^t C_b(u)e^k₂udu + v_bC_b(t), where C, C_b, v_b, K₁, and k₂ are the voxel activity, blood input function, blood fraction, influx rate, and efflux rate. When fitting the kinetic parameters, K₁ is effectively amplified by the factor (1 − v_b)^-1. This works well along the center of the myocardium, where the blood fraction can be as high as 50%; however, for larger blood fractions, such as those encountered within tissue nearer the blood pool, or within infarcted tissue, this correction can amplify K₁ (thus MBF) to unrealistically-high values and diverges as the blood fraction approaches 100%. Thus it is desirable to replace the factor (1 − v_b) with a function that has an inverse which tapers off at blood fractions of 0% and 100%. For this, we propose the Gaussian form 1 − βexp[−(v_b−0.5)²/ 2σ²] , centered on a blood fraction of 50% and parameterized with a suppression factor β and shape factor σ, both of which can be tuned to achieve reasonable flow values. With an appropriate value of σ, the inverse of the Gaussian effectively converges to unity as the blood fraction approaches 0% (pure tissue) or 100% (pure blood); thus it may safely be applied to every voxel without over-correcting. Such an improvement is necessary for the generation of MBF images suitable for clinical interpretation.

Methods: The standard factor (1 − v_b) and the Gaussian factor above were used within the one-tissue kinetic model fit to five Rubidium-82 rest/stress PET/CT scans. The results were compared quantitatively to ensure that the flow values at the center of the myocardium, obtained using the proposed correction, agreed with those obtained using the standard correction. The results were compared visually to assess any improvements due to the proposed correction.

Results: With σ = 0.15 and β = 0.5, the corrected flow values along the center of the left ventricle myocardium, excluding infarcted regions, fall within 2% on average of the flow values corrected using the standard approach. In all cases, visual comparison confirmed the superiority of the proposed method for MBF imaging. Fig. 1 shows a patient stress scan with an apical defect. When applying standard PVC (a), flow values near the blood-tissue boundary are unrealistically amplified. This is resolved using the proposed method (b), resulting in acceptable endocardial boundary delineation and improved visibility of the infarct. Fig. 2 shows a normal patient stress scan, where greater uniformity within the myocardium can be appreciated with the Gaussian PVC.

Conclusions: For MBF imaging, the standard PVC applied to MBF within the one-tissue kinetic model is not effective near the blood-tissue boundary. This can be addressed using the proposed Gaussian PVC. Further studies are needed to fully validate the accuracy of this method for regional assessment of MBF in healthy subjects and diseased patients

• Download figure
• Open in new tab
• Download powerpoint