Abstract
1447
Objectives: Patients with cancer are at increased risk for cardiovascular disease due to common risk factors and anticancer treatments that may lead to toxic effects in the heart. In response to the increasing awareness of the issue of heart health in cancer patients, cardio-oncology is emerging as a new specialty. Whole-body 18F-FDG PET is routinely used for oncological imaging but conventional protocols are not well suited to cardiac imaging. 30-40% of standard oncological FDG-PET scans do not show visible myocardium at all in part because the oncological patient preparation protocol minimizes FDG uptake in heart muscle. In this work, we demonstrate the feasibility of a total-body parametric FDG-PET method using the conventional oncological patient preparation protocol to enable simultaneous visualization and multi-parametric imaging of the myocardium in patients with cancer for integrated cardio-oncological imaging.
Methods: Four consecutive patients with metastatic kidney cancer were injected with 10 mCi of 18F-FDG and underwent a dynamic scan on the EXPLORER total-body PET/CT system. Prior Ethics Committee/IRB approval and informed consent were obtained. Each patient fasted for 6 hours prior to FDG injection. Total-body dynamic data were acquired for 60 minutes and binned into 29 time frames. The static PET standardized uptake value (SUV) was calculated for the last time frame (55-60 minutes). The time activity curve (TAC) from the left ventricle (LV) was used as the image-derived input function for each subject. Two approaches were used for parametric imaging with compartmental modeling. One is the standard approach that uses a single irreversible two-tissue (2T) compartmental model for all image voxels, and the other is an improved approach that deploys model selection with the Akaike information criterion to select the best kinetic model for each voxel from three candidate models: the 2T model, a one-tissue (1T) compartmental model, and a zero-tissue (0T) model (i.e., the blood compartment only). Parametric images of the FDG net influx rate Ki and the blood-to-tissue glucose transport rate K1 were generated for each approach. These parametric images were then compared with the SUV image for visualizing the myocardium.
Results: The myocardium is difficult to see on the standard SUV image in three of the four cancer patients due to either low FDG uptake or lack of contrast with the blood pool. Parametric imaging of Ki using only the 2T model suffers from artificially high Ki values in the blood pool due to overfitting of the model in those blood voxels. Improved parametric imaging with model selection overcomes the overfitting problem by adaptively assigning the appropriate compartmental model to each voxel. The resulting Ki image can improve the visualization of the myocardium. The four patients demonstrate three representative FDG uptake patterns in the myocardium: no or low uptake, diffuse uptake, and focused uptake on the SUV image. Myocardial visualization was improved by Ki (using parametric imaging with model selection) in all subjects. While Ki characterizes glucose metabolism, this method also extracts K1 that reflects glucose transport, thus providing multiparametric imaging of the myocardium in patients with cancer.
Conclusions: Standard SUV images by static PET do not show myocardium in a significant portion of oncological FDG-PET scans. Parametric imaging with voxel-wise model selection can overcome this problem and provides multiparametric imaging of the myocardium without changing the patient preparation protocol. In the future, the parametric images can be further combined with data-driven cardiac gating to measure LV ejection fraction and wall motion. Such a cardio-oncological PET method has the potential to monitor the changes in myocardial glucose transport, glucose metabolism, and cardiac function in patients with cancer for cardio-oncology without the need for an additional dedicated cardiac imaging scan.