| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Marie Curie Research Wing, Mount Vernon Hospital, Northwood, Middlesex, United
Kingdom.
2 Paul Strickland Scanner Centre, Mount Vernon Hospital, Rickmansworth Rd.,
Northwood, Middlesex, United Kingdom HA6 2RN.
3 Siemens Medical Solutions, Forchheim, Germany.
Received May 25, 2005;
accepted after revision August 16, 2005.
Address correspondence to V. Goh
(vicky.goh{at}paulstrickland-scannercentre.org.uk).
Abstract
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
SUBJECTS AND METHODS. Paired perfusion studies were performed on 10 patients who had histologically confirmed advanced non-small cell lung cancer. Using 16-MDCT, multiple sequential volumetric acquisitions encompassing the entire tumor were acquired after infusion of IV contrast material. Using Patlak analysis, median values of tumor permeability (mL/100 mL/min) and blood volume (mL/100 mL) were measured for 10-mm z-axis coverage, and for 40-mm z-axis coverage in each of the paired perfusion studies. Measurement reproducibility was evaluated using Bland-Altman statistics.
RESULTS. Mean difference (95% limits of agreement) for tumor permeability was 1.4 (-4.0 to 6.8) for 10-mm coverage and 0.8 (-3.6 to 5.2) for 40-mm coverage. Mean difference (95% limits of agreement) for blood volume was 1.9 (-5.1 to 8.9) for 10-mm coverage and 1.4 (-3.7 to 6.6) for 40-mm coverage. The coefficient of variation for permeability was 18.7% for 10-mm coverage, improving to 11.9% for 40-mm coverage. The coefficient of variation for blood volume was 41.7% for 10-mm coverage, improving to 32.6% for 40-mm coverage.
CONCLUSION. Our results show that an improvement in tumor perfusion measurement reproducibility may be achieved with greater z-axis coverage.
Keywords: lung • lung diseases • MDCT • neoplasms • oncologic imaging • perfusion CT
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Both dynamic contrast-enhanced MRI and perfusion CT have been used in clinical trials of vascular-modulating drugs to provide such assessment [5-8], although CT has the advantage of being more clinically accessible. Perfusion CT has been validated against a variety of techniques including microspheres, xenon CT, and oxygen-15-labeled H2O PET [9-14] and has been correlated against histologic markers of angiogenesis [15-17]. Until now, CT assessment of tumor perfusion has been limited to a single tumor level with z-axis coverage up to 24 mm. This may be potentially confounding because tumor vasculature is spatially heterogeneous [18, 19]. Greater tumor coverage has the potential to compensate for this heterogeneity and hence to improve measurement variability. This study aimed to determine if measurement reproducibility in lung cancer improves with increasing z-axis tumor coverage by comparing tumor perfusion measurements obtained from 10-mm z-axis tumor coverage with those from 40-mm coverage.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
|
|
|
Using a dual-headed pump injector (Injekttron CT2, Medtron), 100 mL of iobitridol 300 mg I/mL (Xenetix 300, Guerbet) was administered with a decreasing bolus infusion rate (32 mL at 4 mL/s, 16 mL at 2 mL/s, and 60 mL at 1 mL/s) and followed by a saline flush (20 mL at 1 mL/s). The rationale for the contrast-infusion protocol was to optimize conditions for the mathematic analysis model, Patlak analysis, by maintaining a more constant intravascular concentration of contrast material, minimizing the concentration gradient between the intravascular and extravascular spaces, and improving the signal-to-noise ratio during the acquisition.
A single-level bolus tracking scan (CARE bolus, Siemens) at the level of the aortic arch was commenced at the same time as contrast administration using the following parameters: 80 kV; 20 mAs; scanning time, 0.5 seconds; collimation, 4.5 mm; detector width, 0.75 mm. The dynamic study was triggered when peak aortic enhancement was identified from the aortic time-density curve during the bolus tracking scans. The dynamic study consisted of a total of eight breath-hold helical acquisitions, encompassing the entire tumor using the following parameters: 80 kV; 120 mAs; table feed, 30 mm; rotation time, 0.5 seconds; collimation, 2 mm; detector width, 1.5 mm; SFOV, 500 mm; matrix, 512 x 512 mm. Total dynamic acquisition time varied from patient to patient but was approximately 90 seconds. The entire CT perfusion study was repeated within 24 hours without intervening treatment, using identical technical parameters to allow assessment of measurement reproducibility.
Data Postprocessing and Analysis
Data were transferred to a dedicated workstation (Leonardo, Siemens Medical
Solutions). There were a total of 20 perfusion studies (10 patients, two
studies each) for evaluation by a single experienced investigator. Each
perfusion study consisted of nine helical scans (one baseline scan and eight
contrast-enhanced dynamic scans) that required postprocessing before
quantitative perfusion analysis. For each scan, the 2-mm collimated axial
images were reformatted into 10-mm-thick axial images using 3D software (3D
Analysis, Siemens) to permit analysis within a clinically acceptable time.
Reformatted scans were checked to ensure that the whole tumor was included and that each of the reformatted axial images corresponded to a similar position along the z-axis of the patient on all nine scans by comparing the position of the tumor to adjacent anatomic structures. Then each reformatted 10-mm axial image from the same position along the z-axis of the patient from each of the nine helical scans was saved as a separate series on the workstation for further analysis. Thus, for each patient multiple series were obtained for each of the two dynamic studies, encompassing the entire tumor; each series consisted of a single baseline unenhanced image and eight dynamic contrast-enhanced axial images at the same tumor level. The number of series per patient varied depending on tumor size.
For each patient, all series of reformatted dynamic images, encompassing the whole tumor, were loaded into the prototype perfusion software (Siemens) based on Patlak analysis [20]. The arterial input was determined from the bolus tracking scan images for each patient; using an electronic cursor and the mouse, a circular region of interest (ROI) was placed within the aorta. An arterial time-attenuation curve was generated automatically, and this information was saved using the software for subsequent analysis.
A single, central 10-mm tumor level was chosen, and an ROI was drawn freehand around the tumor by a single experienced observer using an electronic cursor and mouse, taking care to exclude surrounding air and atelectatic lung where possible. A tissue attenuation-time curve was generated automatically by the software along with parametric maps of permeability and blood volume (Figs. 1A and 1B). Each pixel location within the functional map corresponded to a single quantitative perfusion value resulting from the mathematic calculation of the data at that location. Data were analyzed on a pixel-by-pixel basis, and median values of tumor permeability and blood volume were derived for this tumor level, equivalent to a 10-mm z-axis coverage. These values were recorded for each patient.
This process was repeated for another three adjacent tumor levels (Fig. 1C). By amalgamating data from all individual pixels from these three levels and the initial tumor level, median values for permeability and blood volume were calculated, producing values for z-axis coverage of 40 mm. Again, these values were recorded for each patient. Analysis of the second perfusion study from each patient was then repeated by the same observer, ensuring similar tumor levels as the first study to allow assessment of measurement reproducibility. Thus, for all 10 patients, median values of tumor permeability and blood volume for 10-mm z-axis coverage and for 40-mm z-axis coverage for both perfusion studies were documented for subsequent statistical evaluation.
Statistical Analysis
Initial analysis was performed to confirm that the statistical assumptions
required for repeatability analysis were upheld. Kendall's tau statistic was
used to establish any relationship between measurement error and the magnitude
of the measurement. If the difference between measurements appeared to
increase when the mean parameter value increased, the data were transformed by
natural logarithm.
Bland and Altman [21-25] statistics were applied to determine the reproducibility between the repeated perfusion studies. The mean difference, SD, and 95% limits of agreement were established. The within-patient coefficient of variation, repeatability coefficient for an individual patient, and variance ratio were also estimated.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
|
|
|
|
|
The mean difference (95% limits of agreement) for tumor permeability was 1.4 (-4.0 to 6.8) for 10-mm coverage and 0.79 (-3.6 to 5.2) for 40-mm coverage. Mean difference (95% limits of agreement) for blood volume was 1.89 (-5.1 to 8.9) for 10-mm coverage and 1.4 (-3.7 to 6.6) for 40-mm coverage. Coefficient of variation for permeability was 18.7% for 10-mm coverage and 11.9% for 40-mm coverage. The coefficient of variation for blood volume was 41.7% for 10-mm coverage and 32.6% for 40-mm coverage. These results show that by increasing z-axis coverage from 10 to 40 mm, an improvement in measurement reproducibility for both perfusion parameters can be achieved.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Single-level techniques have also been problematic for quantitative assessment of perfusion in lung lesions, not least because of image misregistration from respiratory motion during scan acquisition. For example, Miles et al. [26] reported that six of 16 perfusion CT studies of pulmonary nodules could not be analyzed due to image misregistration or beam-hardening artifact from high-density contrast medium within the venous system. To compensate for the spatial variation in tumor perfusion, a solution is to further increase the tumor volume assessed by perfusion CT. By obtaining multiple helical acquisitions dynamically, assessment of a large tumor volume is possible, which may decrease measurement error, thus improving measurement reproducibility. This is particularly important when repeated measurements are made on the same patient, for example, when assessing therapeutic drug effects. Obtaining multiple helical acquisitions dynamically will also reduce problems with image misregistration.
Measurement reproducibility can be evaluated using several indexes. The 95% limits of agreement represents the boundaries within which the true measurement is expected to lie 95% of the time; the narrower the limits, the more precise the measurement being made. The coefficient of variation quantifies the measurement error with respect to the mean and provides an estimation of precision. Both the 95% limits of agreement and the coefficient of variation of our technique are acceptable for response assessment and are comparable to data from the cranial circulation in animals, in which coefficients of variation for cerebral perfusion ranging from 12% to 35% have been reported [27, 28], and data from animal tumor models, in which coefficients of variation for perfusion have ranged from 14% to 24% [9]. More important, an improvement was noted when the z-axis coverage increased from 10 to 40 mm. For example, the coefficient of variation for tumor permeability improved from 18.7% to 11.9%; similarly, limits of agreement narrowed from a range of -4.0 to 6.8 to a range of -3.6 to 5.2. Furthermore, our results compare favorably to single-level dynamic contrast-enhanced MRI reproducibility in human tumors [29], in which the coefficient of variation in log10Ktrans, a measurement of permeability, was 24%.
The repeatability coefficient indicates the 95% confidence limits that might occur spontaneously in an individual. An improvement from 0.5 to 0.3 was shown for tumor permeability when tumor coverage increased from 10 to 40 mm, indicating a decrease in measurement variability. This was also supported by the ratio of between-patient variance to within-patient variance, which increased from 12.4 to 14.2, indicating that perfusion measurements derived from a greater tumor volume may be more sensitive to variations in the parameter studied and less variable when repeated studies are performed on the same individual. This degree of measurement variability is well within the levels of expected therapeutic effect of current antiangiogenic and antivascular drugs undergoing clinical evaluation—for example, bevacizumab (Avastin, Genentech) [5] and combretastatin (OxiGene) (Ng QS et al., presented at the 2003 annual meeting of the European Society for Therapeutic Radiology and Oncology)— but the ability to improve on the reproducibility of this perfusion CT technique bodes well for future clinical utility.
Although we have established that reproducibility of perfusion CT measurements can be improved with increasing tumor coverage, one limitation of this analysis is that it merely provides an overall measure of measurement variability. It does not distinguish among the extrinsic factors that contribute to this variability, including acquisition technique, software variability, and observer variability and intrinsic factors such as tumor heterogeneity. However, on a practical level, identical scanning acquisition parameters and techniques were used in each of the paired studies, and analysis was performed by a single observer using the same software package, minimizing variability from these factors. It is a reasonable assumption that compensation for intrinsic spatial heterogeneity is a major factor contributing to this improvement.
In summary, we have shown that measurement reproducibility improves with increasing tumor coverage. Assessing perfusion over a greater tumor volume may provide more reliable assessment and should be considered in clinical practice. This technique can potentially be used to evaluate whole-tumor vascularity, and further work in developing this technique is ongoing.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
