| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Report |
1 Department of Radiology, Children's Hospital and Harvard Medical School, 300
Longwood Ave., Boston, MA 02115.
2 Department of Radiology, Massachusetts General Hospital and Harvard Medical
School, 32 Fruit St., Boston, MA 02114.
3 Department of Medicine, Massachusetts General Hospital and Harvard Medical
School, 50 Staniford St., 9th Floor, Boston, MA 02114.
4 Harvard—Partners Center for Genetics and Genomics and Harvard Medical
School, 77 Ave. Louis Pasteur, Boston, MA 02115.
Received June 17, 2002;
accepted after revision August 6, 2002.
Supported by contract DAMD 17-98-1-8611 from the United States Army Medical
Research Command.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CONCLUSION. Volumetric measurements of plexiform neurofibromas in neurofibromatosis 1 are reproducible.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The cloning of the neurofibromatosis 1 gene has resulted in insights into pathogenesis that may ultimately lead to the development of nonsurgical therapies. The gene responsible for neurofibromatosis 1 encodes a protein referred to as "neurofibromin," which functions at least in part as a negative regulator of ras family GTPases [4]. Clinical trials that have been undertaken or are contemplated include the use of farnesyl protein transferase inhibitors, angiogenesis inhibitors, cytodifferentiating agents, and hormonal modulators [5, 6].
The high rate of morbidity associated with plexiform neurofibromas makes them a good target for nonsurgical therapies. However, major challenges in the determination of outcomes measurements will complicate these trials. Plexiform neurofibromas may grow erratically, exhibiting periods of rapid growth followed by spontaneous stabilization. Also, the lesions may be large and irregularly shaped, making it difficult to measure their size and to follow changes related to growth or shrinkage in response to treatment.
Given these challenges, and the likelihood that drugs will be available for clinical trial in the near future, we have organized a multicenter trial to determine the natural history of plexiform neurofibromas in neurofibromatosis 1 using volumetric MR imaging. The major goals of that study are, first, to validate volumetric MR imaging as a means of following the growth of plexiform neurofibromas, and second, to generate a body of normative data on the growth rate of plexiform neurofibromas in different regions of the body.
The validity of the volumetric analysis is critically dependent on the ability of an observer to reproducibly determine the margins of a tumor in an MR image. We have studied the interobserver correlation of three observers analyzing volumetric MR imaging data to determine the degree of interobserver variation in the measurement of plexiform neurofibromas and the factors influencing this variation.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The initial part of this study was to compare the volumetric measurements of neurofibromas performed by two radiologists and by a technologist experienced in volumetric assessment of lesions. MR imaging studies of the first 12 consecutive patients recruited were reviewed. All patients had been diagnosed as having neurofibromatosis 1, with the diagnosis established according to clinical criteria. The lesions involved the head and neck (n = 5), spine (n = 4), and trunk or extremities (n = 3).
MR Imaging
Because patients were referred as part of a multicenter study, the MR
imaging units were of different makes and field strengths, with 35 magnets at
1.5 T and one magnet at 1.0 T. The parameters varied according to the area
being examined and the extent of the lesion. Coil and field-of-view selection
varied according to the location of the lesion. The protocol included coronal,
sagittal, and contiguous axial short tau inversion recovery images (TR/TE,
6000/35; inversion time, 150 msec; echo-train length, 8). Axial images were
used for volumetric measurements. Slice thickness was 4 mm for the head and
neck, 5 mm for the spine, and 10 mm for the extremities. The extremities were
imaged with a matrix of 512 x 160 so that the entire lesion could be
covered in a reasonably short time; the other anatomic areas were imaged with
a 256 x 256 matrix.
Volumetric Analysis and Validation with Phantom
The volumetric analysis was performed using Cheshire software (version 4.4;
Parexel, Waltham, MA), a commercially available, Food and Drug
Administration—approved and validated desktop visualization and analysis
program.
Verification of the analysis was performed using a balloon phantom consisting of an elongated balloon filled with 350 mL of water that was wrapped around the lower extremity of an adult volunteer with the intent of simulating a sciatic nerve neurofibroma. This balloon phantom was scanned using axial fast spin-echo inversion recovery technique (6000/35; inversion time, 150 msec; echo-train length, 8). Volumetric measurement was obtained by the computer-assisted method only.
Digital MR images were imported directly to the workstation, or hard-copy images were digitized and imported. Measurements were performed manually. Volume measurements were performed with an Autosegmentation tool of the Cheshire software that determines the best-guess edge of a lesion, based on pixel values, and creates a region of interest (ROI) around the object. On short tau inversion recovery images, neurofibromas are of high signal intensity, whereas solid structures in the body are of very low signal intensity. The segmentation tool therefore successfully identifies the margins of the neurofibromas in most instances. To use this tool, the user must click in the center of the object, drag until the entire circle lies outside the object, and then release the mouse to create the ROI. After the segmentation, the user may use any of the ROI-modifying tools, such as the Nudge tool of the software to adjust the ROI to perfectly outline the object. The Volume Statistics function is used to compute the volume for all areas containing a selected ROI. Standardization of window values is set for all observers between 500 and 600 and the level between 150 and 200.
The technologist was initially instructed by the radiologists on the MR imaging appearance and signal characteristics of neurofibromas, using examples different from the ones used in the evaluation. To perform the reproducibility study, the technologist measured the volume of the lesion on each slice using the Autosegmentation tool. Each radiologist then reviewed the automated measurements and made adjustments using the Nudge tool, according to her or his clinical assessment. The measurements of the two radiologists were made independently and without knowledge of the other's interpretation.
Statistical Analysis
Interobserver reliability was assessed using the intraclass correlation
coefficient, which is the proportion of total variability accounted for by the
variability among subjects. If the intraclass correlation coefficient is high,
then not much of the variability is due to variability in measurements from
different raters [7].
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The results of the measurements are summarized in Table 1 and in Figure 1. In Figure 1, the average of the three observers' measurements is taken as the standard, and deviations are recorded as a percentage of the total volume. As shown on the graph, the variability was usually less than 5%, and in all but two measurements, less than 10%. On the average, radiologist 1 measured a greater volume than the average, and the technologist measured the lowest volume. Measurements from radiologist 1 ranged from 3.2% lower to 19.1% higher than the measurements from radiologist 2 alone. The overall intraclass correlation coefficient was 0.996, which shows excellent agreement among raters. Stratified by group, the intraclass correlation coefficients for the head and neck and for the spine were both greater than 0.999. In Figures 2,3,4, plexiform neurofibromas in three representative locations are shown with the volumetric measurement outlined. The number of sections traced per patient depends on the size of the plexiform neurofibroma. Tracings were saved and recorded.
|
|
|
|
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Determination of the rate of growth of plexiform neurofibromas will be important for selecting tumors appropriate for treatment in clinical trials and for measuring the outcomes of treatment. However, numerous difficulties are encountered in measuring plexiform neurofibromas. In each slice, the lesion may branch in many directions. Unlike most tumors, the borders of neurofibromas are ill defined, and their extension makes adequate lesion coverage challenging. For lesions that are diffuse, ill defined, and large, the time for measuring is longer. In some anatomic areas, it may be difficult to differentiate neurofibromas from normal structures. For example, bowel in the abdomen and pelvis, and lymph nodes in the head, neck, and mediastinum may resemble neurofibromas. Finally, neurofibromas may have different MR imaging appearances in various parts of the body. All these difficulties have created a lack of enthusiasm for or outright skepticism of the possibility of measuring tumor burden in neurofibromatosis 1. The semiautomatic segmentation technique in this study may allow quicker assessment of lesion volume than manual hand tracing.
Our strategy was based on two objectives: to maximize contrast between the lesions and the normal tissues; and to cover the entire lesion. Contrast maximization was best achieved using the short tau inversion recovery sequence with a long TR, which has been used for MR neurography [8, 9]. On long-TR, short tau inversion recovery images, most normal structures in the extremities, head and neck, and spine are of low signal intensity, whereas neurofibromas are of high signal intensity. Slow-flowing vessels are usually indistinguishable from tumor, but we believe that they do not contribute substantially to tumor volume. The technique is less optimal when examining abdominal and pelvic structures because fluid-filled bowel can closely resemble neurofibromas. To cover the entire lesion, we used large fields of view, up to 10-mm slice thickness, and 160 phase-encoding steps, all of which allowed fast imaging, albeit at the expense of optimal image quality. This strategy will likely be equally reproducible in the evaluation of other complex tumoral lesions that have high signal intensity on short tau inversion recovery images.
Our data are limited by a relatively small sample size and the fact that the measurements of the radiologists were not done without their knowledge of the initial assessment by the technologist, which may bias the radiologists. We believe, however, that the major source of disagreement among observers lies with the determination of tumor versus non-tumor on the MR images rather than with the identification of the initial area of interest. We do not have a means of determining the true volume of any of the tumors measured. Our data only address the degree of reproducibility in assessment of tumor volumes as determined by three observers. The data therefore reflect the precision but not the validity of the observations.
It remains to be shown whether the degree of reproducibility will allow detection of growth or shrinkage of plexiform neurofibromas based on serial MR imaging assessment. The current trial of a farnesyl transferase inhibitor, for example, defines progressive disease as an increase greater than or equal to 20% in the volume of the lesion. In this trial, the tumor contours are traced on each image and volume is calculated by summing the results from all images based on the resulting two-dimensional contours and slice thickness. In the ongoing phase II trial, an automated method of tumor tracing for most of the tumors is used to determine the tumor volume. The interobserver variability, usually less than 5% and generally less than 10%, and the high interobserver correlations suggest that determination of volumetric change is feasible. We thus believe that reproducible computer-assisted volumetric analysis of plexiform neurofibromas can be performed successfully, and that such analysis may allow reliable assessment of changes in lesion volume.
Acknowledgments
We thank Tara Flynn, project manager, and Erik Peterson and Noelle
Sittkuhul of WorldCare, Inc., for help in performing volumetric measurements.
We thank Virginia Grove for manuscript preparation.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  K. RADHAKRISHNAN and S. G. ROCKSON The Clinical Spectrum of Lymphatic Disease Ann. N.Y. Acad. Sci., May 1, 2008; 1131(1): 155 - 184. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 E. Dombi, J. Solomon, A. J. Gillespie, E. Fox, F. M. Balis, N. Patronas, B. R. Korf, D. Babovic-Vuksanovic, R. J. Packer, J. Belasco, et al. NF1 plexiform neurofibroma growth rate by volumetric MRI: Relationship to age and body weight Neurology, February 27, 2007; 68(9): 643 - 647. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Theos and B. R. Korf Pathophysiology of neurofibromatosis type 1. Ann Intern Med, June 6, 2006; 144(11): 842 - 849. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
= radiologist 1, 
= radiologist 2, 
=
technologist.
