TO THE EDITOR:
I would like to comment on the viewpoint expressed by Akhurst and Chisin (1) with respect to Wagner’s Newsline article on fused image tomography (2). As a member of the team that designed and built a prototype combined PET and CT scanner (3), I would like to share some of the thinking that led to our choice of clinical PET and clinical CT scanners for the first hybrid device.
Akhurst and Chisin (1) asked if the quality of the CT images should be maximized and, in particular, if the anatomic image should be a clinical-quality CT image. As they pointed out, such a hybrid device provides anatomic images, low-noise transmission scans for attenuation correction of the PET data, and fused functional and anatomic images. They highlighted several drawbacks in offering clinical-quality CT with a hybrid design. These included the difficulty of matching a CT image acquired during a breath-hold to a PET image acquired with the patient breathing normally, the need for subcentimeter registration to answer questions about the exact localization of the PET tracer, artifacts caused by respiratory and cardiac motion that result in incorrect quantitation and image fusion, the requirement for a power injector to administer contrast material, the need to provide a PET technologist trained to perform clinical CT protocols, the demand on the nuclear medicine physician to be trained to recognize anatomic details, and reimbursement issues associated with whether the CT image can be considered the clinical CT image for the patient. They thus suggested that such a hybrid device should combine PET with improved anatomic imaging—improved, that is, compared with current PET transmission scans.
Our prototype design was not foreseen as a PET device with improved anatomic imaging. Instead, we chose to provide full, clinical-quality anatomic images from helical CT, aligned with functional PET images. It is, of course, well known that anatomic and functional images can be acquired on different scanners and aligned retrospectively using software procedures. However, such procedures are not widely used for studies outside the brain because of the technical and logistic difficulties of implementation on a routine basis, even though the advantages of having the CT images available for the interpretation of clinical PET images is well documented (4). The goal of the PET/CT prototype was to overcome some of the technical and logistic difficulties associated with the software approaches by acquiring both clinical-quality anatomic images and clinical-quality functional images in a single scanning session. We feel that such a device will encourage involvement in molecular imaging of other medical specialists such as radiologists, surgeons, and oncologists—specialists who are more familiar with high-resolution anatomic imaging than with the tracer techniques of functional imaging.
The prototype combined PET/CT scanner was designed and built as a collaboration between the University of Pittsburgh and CTI PET Systems, Inc. (Knoxville, TN). The device has been installed in the University of Pittsburgh PET facility since June 1998, occupying a room equipped for the operation of a CT scanner. PET technologists certified in nuclear medicine and with additional training in CT protocols operate the PET/CT scanner, and all scanning is performed with the authorization of the University of Pittsburgh Institutional Review Board governing the use of an investigational device for human studies. A power injector is provided for the administration of intravenous contrast agents during the helical CT scanning, and a radiologist is present during contrast administration. Since the scanner became operational in June 1998, more than 200 patients have been scanned, primarily for oncologic indications. All studies are corrected for attenuation using CT-based attenuation correction factors scaled to account for the difference in energy (5). The studies are read jointly by a radiologist board certified in nuclear medicine and PET-trained nuclear medicine physicians. Additional consultation is available with radiologists specializing in CT of the particular region of the body under examination (e.g., head and neck, thorax, or abdomen).
The ability to perform essentially noiseless transmission scanning and CT-based scatter and attenuation correction is seen as an additional advantage and not as the primary motivation for such a device. Nevertheless, Akhurst and Chisin (1) made a good point concerning the alignment of PET and CT images acquired under different breathing conditions. In this situation, a potential mismatch may compromise the accuracy of both the image fusion and the CT-based attenuation correction factors. Such a mismatch is maximized in the thorax, particularly for the anterior part of the chest wall. Strategies to minimize the mismatch include allowing shallow breathing during both CT and PET scanning, even though the breathing may slightly compromise the diagnostic quality of the CT image. We have conducted several lung studies using this approach, and although some motion artifacts are discernable in the region of the diaphragm and base of the lungs, they generally do not affect the diagnostic quality of the fused image. We have also observed a maximum 5- to 6-mm misalignment of the PET and CT images for tumors in the anterior region of the lungs. Again, such a mismatch does not affect the diagnostic accuracy of the fused image. We have recently begun acquiring CT images with intravenous contrast material for head and neck and oral contrast material for abdominal studies and have verified that contrast-enhanced CT images can also be used for attenuation correction of the PET data. Nevertheless, further validation of CT-based attenuation correction will undoubtedly occur when combined PET/CT scanners become more widely available.
An interesting aspect of the availability of clinical-quality CT images has been the direct participation of radiologists, who raise specific questions related to the location and extent of the FDG uptake and the involvement of adjacent anatomic structures. To answer such questions, precise anatomic localization of the functional abnormality is required. Even with a combined PET/CT scanner, registration accuracy will potentially be affected by patient movement, lesion location, and other study-dependent factors. Therefore, procedures to monitor the quality of the image alignment on a study-by-study basis will be essential to ensure accurate responses to increasingly specific questions.
As Akhurst and Chisin (1) pointed out, good arguments have been made for improving the quality of current PET transmission scans and, at the same time, providing low-resolution anatomic images aligned to functional PET images. However, the prototype PET/CT scanner at the University of Pittsburgh looks beyond that objective to providing radiologists, surgeons, and oncologists the clinical-quality anatomic images with which they are familiar, automatically coregistered with corresponding clinical-quality PET images. Our approach opens up new imaging possibilities and applications and improves a current imaging technology. Admittedly, by potentially crossing existing boundaries, the introduction of such a device into clinical practice will not be simple, straightforward, or rapid. However, if the benefits brought by dual-modality imaging to the management of disease are, as we believe, significant, solutions to issues of technologist training, the availability of nuclear physicians skilled in reading anatomy, and reimbursement will undoubtedly be found. At this early stage in the development of the combined PET/CT scanner, our primary concern has been establishing its role in and contribution to patient care and disease management.
Acknowledgments
This work was supported by National Cancer Institute grant CA65856. David W. Townsend is a consultant for CTI PET Systems, Inc., the company that designed and built the prototype PET/CT scanner.