| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Report |
1
Department of Radiology, FND 202, Massachusetts General Hospital, 55 Fruit
St., Boston, MA 02114.
2
Sarnoff Corporation, 201 Washington Rd., Princeton, NJ 08540.
Received July 12, 2001;
accepted after revision October 2, 2001.
S. L. Aquino was supported in part by the Radiological Society of North
America, and J. C. Asmuth was supported in part by Defense Advanced Research
Projects Agency Medical Initiative under NMA202-97-E-1033-0025.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CONCLUSION. In 11 of 16 data sets of patients imaged for detection of metastatic disease, interpretations from PET studies were correctly altered with registration information. All changes were either improvements in tumor localization or correct interpretation of less metastatic involvement.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
We have applied a computational algorithm using a combined affine and quadratic transformation model to register the CT and PET data sets of patients who were imaged for lung cancer staging or reassessment after cancer therapy. Our goal was to determine whether registered data sets would improve the interpretation of cancer imaging studies, in particular the anatomic detail and physiologic uptake in the mediastinum. We recognize that our transformation model is incapable of accounting for pulmonary nonlinear deformations, and therefore, this work should be considered a step along the way to definitive image registration.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CT Protocol
All patients underwent CT and FDG PET within a 1-month time interval. CT
scans were obtained with HiSpeed or LightSpeed scanners (General Electric
Medical Systems, Milwaukee, WI). Scans were obtained with slice intervals of 5
mm and a pitch of 1:1.5, after approximately 100 mL IV non-ionic contrast
administration. CT scans were acquired during a single breath-hold, with the
patient in the supine position with arms extended above the head.
FDG PET Imaging Protocol
Whole-body and thoracic FDG PET studies were performed with the ECAT-HR+
camera (Siemens/CTI, Knoxville, TN). Image spatial resolution was 5.0-mm full
width half maximum. For PET imaging, the patients were positioned supine on
the imaging bed of the PET scanner with arms at the sides of their bodies or
extended above their heads. Patients fasted at least 6 hr. Blood glucose
levels were measured just before injection of FDG. Approximately 10 mCi (370
MBq) of FDG was injected IV as a bolus. Static emission images, each of 10-min
duration, were obtained beginning about 45 min after injection of FDG. Because
of the limited field of view of the scanner, the patients were imaged in three
contiguous bed positions over the chest. Transmission scans, measured with
rotating rod sources loaded with 68Germanium, were obtained for
each patient. The transmission scans were used for attenuation correction and
for image registration. PET image reconstruction was performed with a
conventional filtered back-projection algorithm.
CT and FDG PET Registration and Image Analysis
Image registration was performed by matching the CT scan with the
reconstructed transmission images. After this procedure, the registration
parameters were used to reslice the PET scans to match the CT images. The
image registration procedure included preprocessing steps to segment the chest
contours, followed by an intensity-based registration algorithm using both
affine and quadratic transformation models. These transformation models
included parameters to account for differences in scale, shear, rotation,
translation, and curvature. Registration was performed using a modified
algorithm [6], which, in a
hierarchic manner, progressed from a coarse registration, with an initially
grossly undersampled volume, to an increasingly refined registration that
included more and more voxels.
All individual PET studies were interpreted by one of two radiologists experienced in nuclear medicine PET, in the light of accompanying CT scans. Registered PET CT data sets were interpreted by the same reviewers, who were unaware of all patient history and prior individual PET interpretation results. Registered data sets were analyzed for areas of increased uptake, and all regions were anatomically specified on the basis of CT anatomy.
Registered images were also analyzed for registration accuracy by visual inspection. Qualitative analysis of fused CT and transmission data sets included alignment of the contours of the lung and mediastinal interface. Evaluation of fused CT and emission data sets included matching of areas of mediastinal uptake on PET with anatomic masses or enlarged lymph nodes on CT, the left ventricular contour, and the esophagus, when visualized.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Comparison of Interpretations of Clinical PET with Registered PET CT
Data Sets
The results of interpretations of registered CT and PET scans and clinical
PET scans of the 15 patients are shown in
Table 1. In patients 2, 3, and
11, clinical PET interpretations could not distinguish radiation changes from
residual tumor. In all three patients, these findings were attributed to
radiation changes on the registered data sets (Fig.
1A,1B,1C).
None of the patients had tumor in these areas on follow-up imaging. In
patients 1 and 6, PET was interpreted as showing areas of increased activity
in the pleura and hila, respectively. These areas were interpreted as normal
on registration. Neither patient had tumor in these areas on follow-up
studies. In patients 7 and 9, unregistered PET was interpreted as showing
mediastinal tumor activity; whereas, after registration, these areas were
shown to be due to gastric activity from a gastric pull-through (patient 7)
(Fig.
2A,2B,2C,2D)
and increased activity at the gastroesophageal junction (patient 9). In
patients 12 and 15, there were differences in anatomic location of increased
nodal activity (Fig.
3A,3B,3C,3D).
In patient 8, unregistered PET was interpreted as showing hilar disease,
whereas the registered data set did not. This patient underwent thoracotomy
and resection, which showed no hilar disease. In patient 14, the description
of nodal activity was more precise. Also in patient 3, an additional focus of
increased FDG uptake was interpreted as increased activity in the lingula,
which could not be identified on CT. The registered data set localized the
area to a mildly thickened pericardium consistent with a pericardial
metastasis. The patient developed a malignant pericardial effusion 4 months
later.
|
|
|
|
|
|
|
|
|
|
|
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
A major obstacle to body image registration arises from the fact that the thorax cannot be considered a rigid body. Issues, such as respiratory and cardiac motion and the deformable nature of the human torso, must be considered to fully address the registration problem. Specifically, CT scans are obtained with the patient supine, with arms extended overhead during a single breath-hold. PET images are usually obtained with the patient in the supine position, during quiet respiration, and with arms often at the side. Accordingly, the body geometry differs between the two scans, making simple rigid body registration unappealing. To properly fuse the two data sets, a nonlinear registration algorithm is needed. Unfortunately, such algorithms have not yet been perfected. Our study uses a registration model that can partially compensate for the confounding issues discussed previously. The experience gained in our study suggests that registration can be useful in assessing the mediastinum.
Prior studies have applied linear registration algorithms to the thorax [7,8,9,10,11]. In studies by Meyer et al. [9] and Magnani et al. [10], an attempt to overcome the issue of patient motion was made by obtaining an additional "nondiagnostic" CT scan with the same imaging parameters as those on PET. In other words, the "nondiagnostic" CT scans were obtained with patients quietly breathing and with their arms down at the sides. These CT data sets were registered to the PET scans. For interpretation purposes, however, an additional "diagnostic" CT was performed with IV contrast administration during a single breath-hold with the arms positioned overhead. This scan was interpreted in accompaniment to the registered data set. The researchers compared the interpretations of individual CT and PET scans with those interpreted together and with registration data sets. Results were compared with surgical pathologic results. They found that the registered data sets were more sensitive and specific in staging primary lung cancer.
In our study, we applied a registration algorithm that obviates obtaining an additional CT scan. This program can register the PET data sets with the clinically acquired helical CT scans. We found that our interpretations from the fused data sets were more detailed and, in some cases, more specific in describing the areas of increased uptake when compared with the PET scans interpreted independently with available CT scans. In seven studies, areas of abnormal uptake attributed to tumor or possible tumor on individual PET interpretations did not show tumor activity. Discrepancies were most often noted in distinguishing areas of increased activity because of radiation changes or physiologic uptake (i.e., esophagus, stomach) in patients who were treated for cancer.
It is essential to anatomically identify changes produced by therapy, whether it is inflammation from radiation therapy or anatomic distortion from combined surgery and radiation. The ability for registered data sets to improve the distinction between recurrent disease or metastasis from therapeutic changes can have a major impact in cancer treatment. In instances in which the studies are indeterminate, ordering physicians must rely on their clinical judgment or follow-up studies to determine if an area of increased activity truly progressed as tumor, resolved as radiation inflammation abated, or remained stable. More precise interpretations would thus help physicians treat and advise their patients in a timely fashion.
PET is already substantially more sensitive than CT in detecting metastatic spread to the mediastinal lymph nodes in patients with primary lung cancer [4, 5]. Prior results of studies evaluating the usefulness of linear registration of PET and CT for lung cancer staging have shown minimal benefit when compared with that of PET interpreted in the presence of CT [11]. Instances in which registration may be most beneficial may exist in cases in which finer anatomic detail is still needed. For example, it is difficult to distinguish lung tumor activity from adjacent hilar or mediastinal activity when the tumor is medially located in the lung. In addition, it is difficult to distinguish hilar node activity from adjacent proximal bronchial and mediastinal activity (Belley et al., meeting of the Radiological Society of North America, Chicago, November 2000). In patients 12 and 15, we described nodal station involvement on registration that was different from that in the clinical PET interpretation. For patient 8, the registered data sets displayed activity only in the upper lobe mass. No activity was identified in the hilum or mediastinum. Clinical PET interpretation, however, described increased activity in the lung mass and the adjacent hilum. At surgical resection, pathology detected tumor only in the lung mass. The hilum and mediastinum were free of disease. In patient 15, who had mediastinal metastases, adjacent hilar nodal disease was described on individual PET interpretation, although none was present on registration. And as we mentioned previously, registration was helpful in displaying uptake in regions of distorted anatomy from prior thoracotomy and radiation therapy.
In conclusion, registration of thoracic CT and PET data sets provided more
specific anatomic localization of areas of increased activity on PET and
accurately downgraded some suspected positive findings on clinical PET
interpretation. We were able to better identify areas of increased activity
from radiation or physiologic changes. This method of further improving the
anatomic detail of already existing tumor imaging techniques may have
potential for improving both radiologic interpretations and
treatment.
,,
|
|
|
Acknowledgments
We thank Daniel Kopans for the use of the Breast Imaging Research
Labaratory and Nathaniel Alpert for his insight and review of this paper.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  M. D. Gilman, A. J. Fischman, V. Krishnasetty, E. F. Halpern, and S. L. Aquino Optimal CT breathing protocol for combined thoracic PET/CT. Am. J. Roentgenol., November 1, 2006; 187(5): 1357 - 1360. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 V. Krishnasetty, A. J. Fischman, E. L. Halpern, and S. L. Aquino Comparison of Alignment of Computer-registered Data Sets: Combined PET/CT versus Independent PET and CT of the Thorax Radiology, November 1, 2005; 237(2): 635 - 639. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 F. C. Detterbeck, J. F. Vansteenkiste, D. E. Morris, C. A. Dooms, A. H. Khandani, and M. A. Socinski Seeking a Home for a PET, Part 3: Emerging Applications of Positron Emission Tomography Imaging in the Management of Patients With Lung Cancer Chest, November 1, 2004; 126(5): 1656 - 1666. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 L. Schrevens, N. Lorent, C. Dooms, and J. Vansteenkiste The Role of PET Scan in Diagnosis, Staging, and Management of Non-Small Cell Lung Cancer Oncologist, November 1, 2004; 9(6): 633 - 643. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Torabi, S. L. Aquino, and M. G. Harisinghani Current Concepts in Lymph Node Imaging J. Nucl. Med., September 1, 2004; 45(9): 1509 - 1518. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
