|
|
||||||||
Technical Developments |
1 German Cancer Research Center, Heidelberg.
| Abstract |
|---|
|
|
|---|
Index terms: Blood, flow dynamics, 20.12114, 57.12113 Computed tomography (CT), contrast enhancement, 20.12114, 57.12113 Computed tomography (CT), diffusion study, 20.12114 Head and neck neoplasms, 20.37, 997.33
| Introduction |
|---|
|
|
|---|
The theoretic concepts of the first technique, which is based on principles of indicator dilution theory and deconvolu-tion analysis, were described by Axel in 1980 (1) and worked out by Berninger et al (2) and Norman et al (3) in 1981. This approach made it possible to determine the mean transit time of a contrast agent bolus and the regional cerebral blood flow in structurally intact brain tissue (ie, in tissue regions without a disrupted blood-brain barrier) (4,5). Most recently, it was also successfully applied to the analysis of dynamic MR images of the brain (69). The application of the indicator dilution theory, however, is limited because contrast media frequently applied in CT and MR imaging studies do not remain within the intravascular space of brain tissues, with disruption of the blood-brain barrier, or in extracranial tissues.
An alternative approach for the analysis of contrast enhancement in extracranial tissues was presented in a pioneering study by Miles in 1991 (10). By adapting a nuclear medicine data processing technique (11), Miles showed that tissue perfusion can be determined from the maximum gradient of the attenuation-time course, measured in the examined tissue region, divided by the peak signal enhancement in the aorta. This relationship is derived under the assumption that no venous outflow and/or extravasation of the contrast medium has occurred before the steepest gradient is determined (12). Although this approach was successfully used to quantify tissue perfusion in the kidney, liver, pancreas, and spleen (10,1214), it should be extended to take into account also bidirectional diffusion of the contrast medium across the capillary wall to allow a more comprehensive characterization of tissue microcirculation.
This is of special interest for the detection and biologic characterization of tumors, since tumor capillaries in general have wide interendothelial junctions, a large number of fenestrae and transendothelial channels formed by vesicles, and discontinuous or absent basement membrane (15). Thus, tumor vessels will display increased permeability for large-molecular-mass substances compared with that in normal vessels (16). Moreover, the interstitial space of tumors is generally much larger than that of normal tissues (17).
In this methodologic study, we therefore present a more comprehensive approach to the analysis of dynamic CT scans, which makes it possible to estimate regional blood flow, capillary permeability, and relative compartmental volumes. This is achieved by detecting and analyzing not only the initial phase of the contrast enhancement but also its further evolution over a period of some minutes.
| MATERIALS AND METHODS |
|---|
|
|
|---|
Patients and Scanning Procedure
Six male patients (mean age, 49.7 years ± 9.7 [SD]) with squamous cell carcinoma of the oro- and/or hypopharyngeal region underwent imaging before combined radiation and chemotherapy. Tumor stage was T4 N2b (n = 2), T4 N2c (n = 2), and T4 N3 (n = 2). The widest diameter of the tumors was between 3 and 8 cm. Written informed consent was obtained from all patients after the nature of the procedure had been fully explained to them. Institutional review board approval has been obtained for our entire study on accelerated hyperfractionated radiation and chemotherapy for patients with advanced cancer of the oro- or hypopharyngeal region, to include treatment and various invasive and noninvasive diagnostic procedures.
After acquisition of three nonenhanced images, 50 mL of a nonionic iodinated contrast medium (iopromide, Ultravist 300; Schering, Berlin, Germany) was administered within 30 seconds via an indwelling forearm canula by using a computer-aided, volumetric pump system (CAI; Doltron, Uster, Switzerland). Starting with the contrast material infusion, sequential CT scans were obtained from the same section with the previously mentioned parameters with a repetition time of 5 seconds for a total of 90 seconds (18 scans) and a repetition time of 15 seconds up to a total scanning time of 6 minutes (18 additional scans). The effective dose per scan was 0.005 mSv (18), resulting in a total effective dose of less than 0.2 mSv per examination.
Image Postprocessing and Data Analysis
Image postprocessing was performed off-line on a personal computer with a Pentium 200 processor under the operating system (Windows-NT, version 4.0; Microsoft, Redmond, Wash) by using IDL (Interactive Data Language, version 3.6; Research Systems, Boulder, Co). For quantitative analysis of the dynamic CT images, regions of interest were defined over the tumor, the carotid artery, and the neck muscles (semispinal muscle of head, splenius muscle of head, or splenius muscle of neck). In general, advanced squamous cell carcinomas of the neck are not well defined at CT; the necrotic and viable parts of the tumors often infiltrate the surrounding muscles and can be hard to distinguish from normal tissue. We, therefore, analyzed only parts of the tumor where the available dynamic and the consecutively measured staging CT scans showed no suggestion of nontumor or necrotic tissue. The muscle region of interest was chosen as large as possible.
The CT attenuation, HT,(in Hounsfield units) of soft-tissue regions was estimated by computing the mean of all pixel values in each region of interest. This approach, however, has been proved to be inadequate for the evaluation of the CT attenuation of blood, HB, from the carotid arteries. Owing to the small size of these vessels, even small patient movements during the measurement period can result in a displacement of the arteries relative to the region of interest. To take this problem into account, we calculated the 90th percentile, instead of the mean, of all pixel values in the region of interest to estimate the actual CT attenuation, HB, in the arteries. This allowed the definition of larger regions of interest, which reduced the probability that the region of interest did not completely coincide with the vessel due to patient movements.
From the CT attenuation values, the mean concentrations of the extracellular contrast agent in the tissue
T and in arterial blood plasma
A (the arterial input function) were calculated from
0.45 is the hematocrit in major vessels, and k is a calibration factor. To improve the signal-to-noise ratio of the curves, they were filtered outside the peak region employing a median filter with a width of three data points.
The pharmacokinetic model employed for the analysis of the concentration-time courses is shown in Figure 1. A list of symbols used in this study is presented in Table 1. According to the open two-compartment model, the transport of a contrast medium through the plasma compartment (with mean contrast agent concentration
P and volume VP) and its diffusion into the interstitial space (with mean concentration
I and volume VI) can be described with the following pair of mass balance equations:
r
1) is always higher than the true plasma flow F. The total tissue concentration
T, which is related to the measurable increase in the CT attenuation according to Equation (1), is given with
|
|
T(t) and
A(t): F/VP, PS/VP, fP and fI. The fifth model parameter is given with PS/VI = PS/VP · fP/fI. The errors in the final fit parameters were computed by means of the covariance matrix.
To be comparable with data presented by other groups, the regional blood volume per unit tissue mass, rBV (in milliliters per 100 g), and the apparent regional blood flow per unit tissue mass, rBF (in milliliters per minute per 100 g), were computed according to
0.25 (21) is the hematocrit in small vessels and m =
· VT is the mass of the soft tissue with mass density
= 1.04 g/cm3 in the examined tissue volume VT. | RESULTS |
|---|
|
|
|---|
As a representative sample of a patient study, Figure 2 shows three of 39 rapid transverse CT images of a dynamic CT data set acquired in a patient with a hypopharyngeal carcinoma.
|
|
|
|
|
|
|
|
|
| DISCUSSION |
|---|
|
|
|---|
The kinetics of contrast enhancement observed in this study with sequential CT scanning in muscle (Fig 3) and tumor tissue (Fig 4) can be described adequately with the proposed compartmental model. The decomposition of the bulk tissue curves into their plasma and interstitial components demonstrates that the initial (perfusion) phase is dominated by the contrast agent in the intravascular compartment (permeabilitylimited transport, PS/F < 1) and, thus, can be used to estimate tissue perfusion in a reasonably good approximation as, for example, using the steepest-gradient technique. However, if the contrast enhancement is not permeability limited, the steepest-gradient technique may result in a marked overestimation of blood flow. Moreover, the decomposition of the contrast-enhancement curves reveals that the intravascular component cannot be neglected, as is the case in the previously mentioned simple permeability models. The differences in the tumor enhancement profiles shown in Figure 4b and 4d, for example, are mainly due to differences in the intravascular component, whereas the interstitial contribution is nearly identical.
To roughly examine the validity of our approach, attenuation-time courses were measured in normal neck muscles of six patients and the regional blood volume and regional blood flow values derived were compared with results obtained by other groups with various measurement techniques (Table 2). The regional blood volume value of 4.9 mL/100 g ± 1.3 determined with our CT technique is slightly higher than the literature values, which are in the range between 1.8 mL/100 g ± 0.1 and 4.7 mL/100 g ± 0.6. This may be explained (a) by the fact that the neck musclesin contrast to the muscles of the leg or forearmwere not at rest since the patients were instructed not to move their heads during the examination period and (b) by a small vasodilatory effect of the employed nonionic contrast medium (33,34). The relative fractions of the plasma (fP = 0.04 mL/mL ± 0.01) and interstitial tissue compartment (fI = 0.12 mL/mL ± 0.01) also coincide well with scarce literature data available for muscle tissue. For skeletal muscle of the rat, O'Connor and Bale (35) reported a relative fraction of the interstitial volume of 0.14 mL/mL ± 0.02 and Gullino et al (36) a relative volume of the extracellular space (
fP + fI) of 0.16 mL/mL ± 0.01.
The muscle blood flow of 12.0 mL/min/100 g ± 5.4 estimated in our study, on the other hand, is considerably higher than the data obtained by other groups, varying between 3.1 and 7.0 mL/min/100 g, but is still smaller than the blood flow value of 13.5 mL/min/100 g ± 3.3 determined with positron emission tomography (PET) during ergometric exercise of the calf muscle (Table 3). This discrepancy is mainly due to the fact that the apparent plasma flow F = F/r (0
r
1), which can be estimated from the measured CT enhancement profiles according to our model, is considerably higher than the true plasma flow F. With the assumption, for example, that the concentration of the contrast medium in the capillaries decreases linearly between the inlet and the outlet of a capillary, it follows that r = 1/2 (Appendix), which means that the apparent flow F is twice as high as the true flow F. However, it should be recognized that the other three model parameters (fP, fI, and PS/VP) are not affected by this systematic overestimation.
The surface area of the capillaries per unit volume of skeletal muscle tissue, S/VT, is known to be approximately 130 cm2/cm3 ± 25 (mean ± SEM calculated from the results of different studies summarized in a review by Jain et al [16]). From this, the mean permeability of muscle capillaries for the contrast agent iopromide (molecular mass, A = 791 d) can be estimated to be P = fP · (PS/VP)/(S/VT) = (0.8 ± 0.4) · 10-5 cm/sec. This value fits well to permeability coefficients found in muscle tissue for Na+ (P = 5.1 x 10-5 cm/sec, A = 11 d), sucrose (P = 1.4 x 10-5 cm/sec, A = 342 d), and inulin (P = 0.2 x 10-5 cm/sec, A = 5,000 d) (16).
The pharmakokinetic parameters summarized in Table 3 demonstrate that both the plasma and the interstitial spaces in the examined neck tumors are much larger than those in muscle tissue (fP = 0.13 mL/mL ± 0.02 vs 0.04 mL/mL ± 0.01, fI = 0.27 mL/mL ± 0.02 vs 0.12 mL/mL ± 0.01). The estimated tumor volumes fit well into the range of data summarized in a comprehensive review article by Jain (17) for a variety of experimental sarcomas and carcinomas in the rat (fI = 0.330.60 mL/mL), as well as for different human brain tumors (fP + fI
0.130.40 mL/mL).
Results of a few studies of blood flow through squamous cell carcinoma of the head and neck region in humans have been reported. Data published until 1992 are summarized in a review article by Vaupel and co-workers (37). Despite similar histologic classification and primary site, the measured blood flow values vary considerably between 6 and 25 mL/min/100 g. Compared with these data, the mean apparent blood flow determined in the present study of 45 mL/min/100 g ± 16 (range, 2263 mL/min/100 g) is considerably higher due to the reasons discussed previously. It is interesting to note, however, that the tumor-to-muscle blood flow ratio of 3.8 ± 2.2 obtained in our study is very close to the tumor-tonormal tissue ratio of 3.9 ± 2.7 determined by Wheeler and co-workers from washout curves of intraarterially injected xenon-133 (38). Relatively high blood flow values in squamous cell carcinoma of the head and neck were also reported by Hermans et al in 1997 (39). They estimated tissue perfusion from dynamic CT scans by means of the steepest-gradient technique in 18 patients and found a mean blood flow of 75 mL/min/100 g, varying between 28 and 132 mL/min/100 g.
As an interesting result, the higher regional blood flow observed in our study in the neck tumors (45 mL/min/100 g ± 16) in comparison with that in muscle tissue (12 mL/min/100 g ± 5) is mainly due to the higher regional blood volume and not to a higher flow per unit plasma volume F/VP (Eq [6]), which is nearly identical for tumor (2.7 mL/mL/min ± 0.8) and muscle tissue (2.5 mL/mL/min ± 0.9). The surface-area product per unit plasma volume (PS/VP) was also found not to be higher in the neck tumors than in muscle tissue. This is consistent with the findings of Jain (40) and Gerlowski and Jain (41) that the permeability of compounds with a molecular mass of less than 2,000 d is almost identical in normal and neoplastic tissues. As pointed out by different investigators, this makes a strong argument for the use of contrast media with large molecular mass for the assessment of abnormal capillary permeability and thus for improved tumor detection. The pharmacokinetic properties of conventional contrast agents, on the other hand, make them an ideal probe for studying the chances of circulating low-molecular-mass compounds (eg, anticancer drugs) to distribute into the interstitial space of defined tissues. This is particularly interesting for solid malignant tumors as, in this case, the delivery of systemically administered drugs to the target site is one of the major problems of drug therapy.
The tumor parameters summarized in Table 3 indicate that the relative compartmental volumes fP and fI can be determined with a greater accuracy and, moreover, show a smaller interindividual variation than the functional parameters characterizing the transport and extravasation of the applied contrast medium (F/VP and PS/VP). Whereas the first observation mainly reflects the numeric robustness of the nonlinear parameter estimation against measurement errors, the second observation reveals that functional variability between tumors is larger than are structural or histologic variability. This is in accordance with all available information on parameters characterizing microcirculation in human cancer. According to these data, marked variations in physiologic properties are likely to occur between tumors with the same grade and clinical stage and, moreover, between different locations within a tumor (42). Owing to this reason, care must be exercised when comparing parameters obtained for tumors in different individuals (16). Despite the fact that the physiologic tissue parameters estimated in the present study agree well with published data, it will thus be necessary in future studies to experimentally validate our perfusion-permeability model in more detail on an individual basis. In an animal study, for example, CT measurements of blood flow in normal tissues and in transplanted tumor models can be correlated with flow values determined by means of simultaneously injected radiolabeled microspheres. Moreover, the relative volume of the intravascular and the extracellular space can be estimated by means of gamma counting of tissue samples after in vivo administration of a radiotracer confined to blood vessels and of another one excluded by cells.
Independent of whether our model can be validated with such experiments in all details, it offers promising prospects for the clinical radiologist to quantitatively characterize tissue microcirculation. It is, for example, a fundamental concept in oncology that tumor growth and metastatic potential depend on the ability of a tumor to induce formation of new microvessels (4345), often with severe structural abnormalities (15,16,46). It is therefore speculated that monitoring of tumor neovascularization might make it possible to predict tumor aggressiveness (eg, risk of lymphatic involvement and disease-free survival after therapy) and to design an individually tailored therapeutic approach (45,4751). As a result of a compromised and heterogeneous microcirculation, many human malignancies comprise hypoxic tissue areas that may lead to a failure of conventional radiation therapy (46,52). Such hypoxic tumors need to be identified and selected for special procedures that exist for dealing with hypoxic cells but cannot be employed in all patients (39). Since tissue perfusion is correlated with tissue oxygenation, individual radiation therapy planning might be possible in the future on the basis of physiologic parameters determined with noninvasive sequential CT.
The pharmacokinetic model derived in this study can also be employed to analyze dynamic MR measurements for administration of a diffusible paramagnetic MR contrast agent (53). Findings in first applications in dynamic MR mammography showed that the large variety of enhancement patterns observed in different breast tissues (eg, glandular tissue, mastopathic nodules, fibroadenomas, carcinomas) can adequately be parameterized with our perfusion-permeability model. Up to now, however, it is an open question as to what extent the diffusive behavior of water protons or the variation of proton relaxivities in different tissue compartments affect the reliability of functional tissue parameters determined from dynamic MR images (54). Since there are MR and CT contrast media with similar physicochemical properties, it may be possible to answer these questions with a comparative analysis of dynamic CT and MR data sets acquired in the same patients.
One limitation of the dynamic CT technique is undoubtedly its radiation exposure for the patient. Despite the fact that the dose per scan was reduced as far as possible in our study (25 mAs) the total effective dose may still be too high for a diagnostic procedure under certain circumstances. It should be recognized, however, that a comparable extensive quantitative characterization of tissue microcirculation can otherwise be obtained only with multitracer PET studies, which lead to a markedly higher radiation exposure and, moreover, in general make it necessary to take arterial blood samples in short time intervals to determine the arterial input function of the applied tracers.
In clinical practice, one may also be faced with the problem that dynamic CT scans can be acquired with a sufficiently high temporal resolution from only one section. It may thus be difficult in certain cases to determine the arterial input function, since a large vessel is not in the field of view. With the advent of a new generation of CT scanners with multirow detectors, however, it will be possible in the near future to determine blood and tissue curves from different sections.
In conclusion, the proposed pharmacokinetic analysis of dynamic CT data offers promising prospects to quantitatively assess tissue microcirculation with physiologic parameters, concretely with the regional blood flow, the capillary permeability, and relative compartmental volumes. Although our approach is not yet validated in all details, the agreement of our results with published data indicates that it may be employed in patient studies to improve our understanding of tissue pathophysiology and of the action and optimization of therapeutic interventions.
| APPENDIX |
|---|
|
|
|---|
One approach was presented by Kety (56) for inert gases that are rapidly distributed in the interstitial and intracellular spaces. Since the capillary volume represents less than 5% of the volume in most tissues, it is reasonable to assume that changes in the mean tissue concentration of the tracer during the time of passage of an element of blood through the tissue is negligible in comparison with the concentration change occurring in the blood itself. By using this postulate, Kety derived an analytic solution for the total amount of tracer in a tissue region from which the transfer coefficient K = F · E/Ve can be determined with Ve the partition volume of exchangeable tracer and E = 1 - e-PS/F the extraction fraction. Whereas the Kety model can also be applied to activity-time courses measured with PET after the administration of oxygen-15labeled water, it is not applicable to extracellular contrast media usually employed in CT and MR imaging studies. In contrast to inert gases or O-15labeled water, these contrast media are distributed within only the interstitial volume, which represents less than about 20% of the total volume in tissues. Thus, changes in the mean tissue concentration of the tracer during the passage of an element of blood cannot be ignored.
An alternative approach was presented by Morales and Smith (57). These authors assumed only that the difference CA -
P between the arterial and the mean capillary contrast agent concentration is some constant fraction 0
r
1 of the difference CA - CV between the arterial and venous concentrations, that is,
F describing the apparent plasma flow. A graphic visualization of the mathematic model defined with the two coupled differential Equations (A6) and (A4) for
P and
I is given in Figure 1. In contrast to the Kety model, the approach formulated by Morales and Smith makes it possible to independently characterize plasma flow and diffusion across capillary endothelium. The rate of these transport processes is described with the apparent plasma flow F and the permeabilitysurface area product PS, respectively. | Acknowledgments |
|---|
| Footnotes |
|---|
Author contributions: Guarantor of integrity of entire study, G.B.; study concepts, G.B., W.S.; study design, G.B., M.L.B.; definition of intellectual content, G.B., M.L.B.; literature research, G.B.; clinical studies, M.L.B.; data acquisition, M.L.B.; data analysis, G.B., U.H., A.H.; manuscript preparation and editing, G.B.; manuscript review, M.L.B., U.H., A.H., W.S.
Received January 28, 1998;
revision requested March 6, 1998; revision received May 21, 1998;
accepted July 20, 1998.
| References |
|---|
|
|
|---|
This article has been cited by other articles:
![]() |
K Murase, S Miyazaki, and X Yang An efficient method for calculating kinetic parameters in a dual-input single-compartment model Br. J. Radiol., May 1, 2007; 80(953): 371 - 375. [Abstract] [Full Text] [PDF] |
||||
![]() |
J. M. Provenzale, S. Mukundan, and D. P. Barboriak Diffusion-weighted and Perfusion MR Imaging for Brain Tumor Characterization and Assessment of Treatment Response. Radiology, June 1, 2006; 239(3): 632 - 649. [Abstract] [Full Text] [PDF] |
||||
![]() |
J. C. Miller, H. H. Pien, D. Sahani, A. G. Sorensen, and J. H. Thrall Imaging Angiogenesis: Applications and Potential for Drug Development J Natl Cancer Inst, February 2, 2005; 97(3): 172 - 187. [Abstract] [Full Text] [PDF] |
||||
![]() |
L. H. Dennis Cheong, C. C. Tchoyoson Lim, and T. S. Koh Dynamic Contrast-enhanced CT of Intracranial Meningioma: Comparison of Distributed and Compartmental Tracer Kinetic Models--Initial Results Radiology, September 1, 2004; 232(3): 921 - 930. [Abstract] [Full Text] [PDF] |
||||
![]() |
K L Li, X P Zhu, D R Checkley, J J L Tessier, V F Hillier, J C Waterton, and A Jackson Simultaneous mapping of blood volume and endothelial permeability surface area product in gliomas using iterative analysis of first-pass dynamic contrast enhanced MRI data Br. J. Radiol., January 1, 2003; 76(901): 39 - 51. [Abstract] [Full Text] [PDF] |
||||
![]() |
J.-P. Delille, P. J. Slanetz, E. D. Yeh, D. B. Kopans, and L. Garrido Breast Cancer: Regional Blood Flow and Blood Volume Measured with Magnetic Susceptibility-based MR Imaging—Initial Results Radiology, March 21, 2002; (2002) 2232010428. [Abstract] [Full Text] |
||||
![]() |
A. F. DeVries, J. Griebel, C. Kremser, W. Judmaier, T. Gneiting, A. Kreczy, D. Öfner, K.-P. Pfeiffer, G. Brix, and P. Lukas Tumor Microcirculation Evaluated by Dynamic Magnetic Resonance Imaging Predicts Therapy Outcome for Primary Rectal Carcinoma Cancer Res., March 1, 2001; 61(6): 2513 - 2516. [Abstract] [Full Text] |
||||
![]() |
B. E. Van Beers, I. Leconte, R. Materne, A. M. Smith, J. Jamart, and Y. Horsmans Hepatic Perfusion Parameters in Chronic Liver Disease: Dynamic CT Measurements Correlated with Disease Severity Am. J. Roentgenol., March 1, 2001; 176(3): 667 - 673. [Abstract] [Full Text] [PDF] |
||||
![]() |
J.-P. Delille, P. J. Slanetz, E. D. Yeh, D. B. Kopans, and L. Garrido Breast Cancer: Regional Blood Flow and Blood Volume Measured with Magnetic Susceptibility-based MR Imaging—Initial Results Radiology, |