Current state of hybrid imaging: attenuation correction and fusion

Abstract

Hybrid positron emission tomography (PET) and computed tomography (CT) scanners were introduced in the late 1990s. Hybrid imagers were first used for oncology and were widely accepted for their unique ability to merge high-resolution anatomy with F-fluorodeoxyglucose (FDG) as a functional tumor agent. For cardiac applications, the CT data was useful only for attenuation correction, as the motion of the heart caused significant artifacts in the thoracic images of the mediastinum from the single slice CTs in these early scanners. Multidetector CT scanners for high-resolution cardiac imaging were integrated into hybrid systems starting in about 2002. For the first time it appeared that high-resolution cardiac anatomy would be easily correlated to myocardial perfusion images (MPI) using the same imager to acquire both images in a single scanning session. However, it soon became apparent that thoracic CTs obtained from hybrid scanners were not necessarily in registration with the serially acquired PET, and some adjustments had to be made for these differences. Misalignment of emission and x-ray CT images is a particular challenge in the thoracic cavity due to motion of the diaphragm, heart, abdominal organs, and bony structures, which are not consistently represented in the two studies. In attenuation correction, breathing motion is generally considered the main problem; transmission and emission scan misalignment is most noticeable at the left lung/left-ventricle interface where myocardial uptake overlies the left lung of the CT on fused images as well as at the lung/diaphragm interface where tissue in the transmission scan may or may not be present at the lung/diaphragm interface of the emission. When considering fusion for correlating anatomy and physiology, breathing motion is definitely a problem, but cardiac motion is concerning for additional reasons. Generally, coronary arteries are observed best on only a few frames of the dynamic computed tomographic angiogram (CTA) study (and in some protocols, only one frame is acquired). The nuclear study is generally summed to improve signal-to-noise prior to visual or quantitative analysis. Thus, there is an inherent mismatch between the hearts as imaged in the PET and CT. In addition, the first attempts to display these images simply used overlay techniques, similar to what was being done for oncology. However, the correlation of the perfusion information with the coronary anatomy could not be properly appreciated in a slice-by-slice display, or could quantitation of the perfusion studies be easily included. To address the challenge of correcting transmission and emission misalignment for attenuation correction, several groups have suggested solutions including breathing instructions during the CT acquisition, modifying the CT protocol to allow for free-breathing, and retrospective or post-processing registration techniques. For fusion of CTA and MPI, special alignment techniques have been developed for registering the two images, since nearly perfect alignment is important for diagnostic purposes. Given that such post-acquisition registration software may be needed for hybrid systems, investigation into its use for aligning studies acquired from separate machines is also being performed. Finally, high resolution displays showing 3D coronary anatomy concurrently with 3D perfusion information, including quantitative analysis of perfusion defects, are being perfected and their importance for CAD diagnosis is being demonstrated. These topics are addressed in the following sections.