DOI: 10.1148/radiol.2372041403
(Radiology 2005;237:670-674.)
© RSNA, 2005
Small-Bowel Perfusion Measurement: Feasibility with Single-Compartment Kinetic Model Applied to Dynamic Contrast-enhanced CT1
Arkadiusz Sitek, PhD and
Robert G. Sheiman, MD
1 From the Department of Radiology, Beth Israel Deaconess Medical Center and Harvard Medical School, 330 Brookline Ave, Boston, MA 02215. Received August 13, 2004; revision requested October 13; revision received November 22; accepted December 24.
Address correspondence to R.G.S. (e-mail: rsheiman{at}BIDMC.harvard.edu).
 |
ABSTRACT
|
|---|
This study was institutional review board approved and HIPAA compliant. Informed consent was obtained from all patients. The purpose of the study was to prospectively examine the feasibility of measuring small-bowel quantitative blood flow by using motion-corrected, contrast-enhanced computed tomographic (CT) images and a single-compartment kinetic model. Seven patients underwent abdominal CT in which 40 10-mm-thick sections were obtained at a single level. Small-bowel images were obtained every 3 seconds after contrast agent administration. Automated application of regions of interest yielded time-enhancement curves for the bowel wall and the aorta. A one-compartment model was applied to each set of time-enhancement curves for determination of the small-bowel volumetric blood flow FV, volume of distribution VD, and blood transit time 
. FV was also calculated by using the first-pass method and 
variate analysis for model validation. The FV values obtained by using the single-compartment model (mean FV, 0.47 min–1) showed excellent linear correlation with those obtained by using the first-pass method (Pearson r = 0.80) and variate analysis (Pearson r = 0.97). Mean VD and values were 2.86 (unitless) and 4.28 seconds, respectively. A one-compartment kinetic model can be applied to motion-corrected, contrast-enhanced small-bowel CT images to quantify perfusion.
© RSNA, 2005
|
INTRODUCTION
|
|---|
Computed tomography (CT)-based tissue perfusion techniques are based on the simple and known direct correlation between blood enhancement after intravenous contrast agent administration and blood volumetric flow. Many methods of tissue perfusion have been proposed and are now routinely used. Perfusion imaging based on the deconvolution method introduced by Axel (1) has been successfully used in brain CT imaging (2,3), while perfusion imaging of normal and diseased livers has been performed by Miles et al (4,5) by using a first-pass method (FPM). Variate analysis represents another approach to studying tissue perfusion at CT and has been used to model arterial enhancement and thus perfusion (6). CT software for perfusion analysis is currently available and is based on either a deconvolution method or variate analysis.
Deconvolution methods are based on the assumptions of a uniform distribution of tracer (iodinated contrast material for CT) within tissue and a shape of the tissue's impulse response—that is, how a molecule of contrast material responds while it is within tissue. Other methods of assessing perfusion involve the use of only the initial upsloping portion of the time-enhancement curve (TEC) of the target tissue, either neglecting (with the FPM) or underestimating (with variate analysis) the remainder of the curve (6).
To our knowledge, CT-based assessment of small-bowel perfusion had not been undertaken before now, probably because accurately applying a region of interest (ROI) to the small-bowel wall to generate a TEC is difficult and undermined by motion. However, quantitating small-bowel perfusion may be useful because the small-bowel blood flow at Doppler ultrasonography has been shown to correlate with the level of inflammatory bowel disease activity (7,8) and from a logical standpoint should be altered in multiple disease states, most notably ischemia. Therefore, the purpose of our study was to prospectively examine the feasibility of measuring small-bowel blood flow by using motion-corrected, contrast-enhanced CT images and a single-compartment kinetic model.
|
Materials and Methods
|
|---|
Patients
This study was Health Insurance Portability and Accountability Act compliant and approved by our institutional review board; informed consent was obtained from all patients. The seven consecutive patients (four men, three women; mean age, 63 years ± 16 [standard deviation]) who constituted our study population underwent diagnostic CT without oral contrast material for evaluation of potential genitourinary or pelvic abnormalities. Absence of known or suspected gastrointestinal abnormalities was the only inclusion criterion and was confirmed by means of medical record review and interviews with the patients at the time they gave informed consent. All seven patients met this criterion.
All patients underwent bowel perfusion CT, which involved the acquisition of 40 10-mm-thick, 36-cm-field-of-view transverse images every 3 seconds. Because imaging was performed for 120 seconds to ensure the acquisition of data during all phases of perfusion (ie, contrast material wash-in and washout), the patients were placed in the prone position and performed shallow breathing to minimize respiratory-induced motion. To avoid the use of a test bolus of contrast material, which theoretically may affect the TEC owing to recirculation, we neither determined nor implemented a time delay in image acquisition. Also, the ideal approach to modeling any system mathematically is to assess the system's response to a short, intense input. With this in mind, parameters that would facilitate the administration of a short, intense bolus of contrast material in the small bowel were chosen. Thus, imaging began simultaneously with the intravenous injection of ioversol (Optiray 320; Mallinckrodt, St Louis, Mo) at 4 mL/sec through an antecubital vein, for a total injected dose of 40 mL.
All images were obtained at the same z position, without table motion, at a low radiation dose (100 kV, 80 mA; estimated total dose, 7.74 mSv), and by using a four–detector row CT scanner (LightSpeed; GE Medical Systems, Milwaukee, Wis). To allow a z position that encompassed the proximal jejunum in the transverse plane and the abdominal aorta at the level of the superior mesenteric artery, the table position was chosen after transverse CT images of the abdomen were obtained. For bowel preparation, the patients drank approximately 36 oz of water 30 minutes before undergoing scanning to facilitate bowel distention and wall delineation. No medication was given in an attempt to decrease bowel peristalsis to avoid any potential effects on bowel perfusion.
Image Analyses
TECs of the small bowel were obtained in the following manner: For consistency, all ROI measurements were a minimum of 500 pixels (25 mm2) and were performed on the proximal jejunum just beyond the expected location of the ligament of Treitz. We made an effort to choose a small-bowel loop that was oriented perpendicular to the transverse plane to ensure transmural depiction. We developed dedicated software that differed from commercially available software in that it enabled retrospective placement of an ROI of any chosen size at a specific location on a single transverse image and then allowed the same ROI to be reproduced on all subsequently obtained images at the identical location. The software instantaneously subtracts the enhancement value (in Hounsfield units) for the first image (assumed to be nonenhanced) to yield a TEC, a tabulated presentation of the ROI enhancement at each time point, and associated minimum and maximum enhancement values. All initial TECs were created by a single investigator (A.S.). An additional feature of this software is that it enables the ROI position to be adjusted on a single image (time point), with instantaneous updating of the TEC and the tabulated data for that time point only.
Because peristalsis and respiratory motion may cause the applied ROI to move in relation to the targeted bowel, after initial TEC generation, each image obtained in a given patient was reviewed by the same investigator to ensure accurate positioning of the ROI on bowel wall only. Adjustments were made when (a) an ROI had a minimum enhancement value of 0 or less, indicating the inclusion of pixels that were either overlying the bowel lumen (containing water) or beyond the bowel wall (containing fat) or (b) it was subjectively believed that the ROI included an enhancing structure adjacent to the bowel wall (eg, a small vessel). Bowel motion was corrected for in this manner. The number of adjustments performed for each subject was noted. A similar process was performed to generate a TEC for the abdominal aorta (Fig 1).
View larger version (97K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1. Creation of a TEC with software designed for this study. An ROI (circle) was placed within the aorta on a single transverse contrast-enhanced CT image obtained with the patient in the prone position. The ROI was automatically reproduced on all 40 images at the identical location to yield an aortic TEC (green). The red dot corresponds to the image currently shown. Clicking on any single time point prompts the display of the image associated with that point, with the TEC superimposed. Any adjustment of the ROI position on an image can be applied to only that image or to all images, with the TEC automatically updated and redisplayed. Actual updated enhancement values and the corresponding times will also appear on the same screen (not shown).
|
|
Blood Flow Parameters
Small-bowel blood flow parameters were determined by applying a one-compartment kinetic model to mimic small-bowel perfusion (Appendix). Like other methods of evaluating tissue perfusion, our model made two basic assumptions: (a) that there was uniform mixing of contrast material in the blood at the level at which our TECs were generated such that the enhancement changes that occurred with time directly correlated with the change in iodinated contrast material concentration and thus with the change in blood flow and (b) that the transfer of iodinated contrast material to the extravascular space was negligible. The aortic TEC is considered to be identical to that of an arteriole feeding bowel (contrast material concentration does not change as it traverses branching arteries, as per the work of Clough et al [9]) and represents Cin(t), while the bowel TEC represents C1(t). This model makes no assumptions concerning impulse response and makes use of the entire TEC, and, thus, it also considers the outflow or washout of contrast material from the bowel. The following differential equation is derived from the kinetic one-compartment model and results from a mass balance of contrast material (iodine) entering the small bowel:
 | (1) |
where FV represents the blood flow per unit volume of small bowel tissue, or the small-bowel volumetric blood flow, in milliliters per minute per milliliter (ie, minutes–1). VD is the volume of distribution (unitless) and is the reciprocal of the proportion of a unit volume of bowel that is available to iodine and thus blood—that is, the model does not assume uniform distribution of contrast material within the bowel. represents the time delay, or blood transit time, in seconds, between the time at which the contrast material arrives in the aorta at the superior mesenteric artery level and the time at which it reaches the targeted small bowel; this value is an intuitive indicator of the velocity of blood traveling to the bowel. The nonlinear Marquardt least-squares method (10) was used to solve Equation (1) for each patient, where C1(t) and Cin(t) are known values represented by points on the bowel and aortic TECs, respectively, at a given time (t). By solving Equation (1), the optimal values of , FV, and VD for each patient can be determined. The optimal values of FV, VD, and for each subject were then used to generate modeled small-bowel TECs, which were compared with the actual measured TECs for wellness of fit.
Because, to our knowledge, no model had been applied to contrast-enhanced CT images for assessment of small-bowel perfusion before now, and because a one-compartment kinetic model yields and VD values, which are not obtainable with use of the current commonly used techniques, we believed that it was necessary to assess the relative validity of our one-compartment kinetic model. Therefore, for each subject, the FV value obtained by using the kinetic model was also calculated by using two accepted techniques for assessing tissue perfusion: variate analysis and the FPM. For each small-bowel wall TEC, a variate was fit by using a least-squares technique, while the FPM involved the use of the maximum rate of bowel enhancement relative to the peak enhancement of the arterial inflow into bowel:
 | (2) |
where Etis is the maximum rate of bowel tissue enhancement, in Hounsfield units per minute, and Eart is the peak arterial enhancement, in Hounsfield units. The FV values obtained by using both of these accepted methods were compared with the optimal FV for each patient obtained by using the one-compartment kinetic model. We did not compare the FV values obtained by using variate analysis and the FPM.
Statistical Analyses
Pearson correlations were used to compare the seven FV values derived from the kinetic model with those obtained by using variate analysis and the FPM. The Pearson correlation coefficient (r) reflects the linear relationship between the FV values obtained by using each method for each patient, has a maximum value of 1.0, and indicates strong correlation when the r value is 0.8 or greater. Each model-derived small-bowel TEC was compared with the corresponding measured TEC by calculating the maximum regression value (R2), where R2 has a maximum value of 1.0, which indicates a perfect fit between the modeled and measured TECs. Analyses were performed by using computer software (Data Analysis Toolkit, Excel 2002; Microsoft, Redmond, Wash).
|
Results
|
|---|
The optimal values for , FV, and VD for all seven patients derived from our kinetic model are shown in the Table, and the R2 values derived from comparisons between each modeled TEC and each measured TEC are illustrated in Figure 2. Overall, the model yielded an excellent fit to the measured TECs, with all R2 values corresponding to P < .001 and ranging from 0.80 to 0.98. The mean number of small-bowel ROI adjustments needed per patient to achieve a TEC was four (10% of the measurements).
The FV values obtained by using the one-compartment model ranged from 0.31 to 0.78 min–1 (mean, 0.47 min–1 ± 0.18 [standard deviation]). The FV values obtained by using the FPM and variate techniques showed a strong linear correlation with our model: The mean FV with the FPM was 0.50 min–1 ± 0.37 (r = 0.80, P < .05), and the mean FV with the variate analysis was 0.35 min–1 ± 0.16 (r = 0.97, P < .001). The mean small-bowel VD was 2.86 ± 0.6, and the mean was 4.28 seconds ± 1.4.
|
Discussion
|
|---|
Our data indicate that small-bowel TECs can be constructed by using motion-corrected CT images, and, more important, that small-bowel perfusion parameters can be measured by using these corrected contrast-enhanced CT images. Given that we determined FV values similar to those obtained by using both variate analysis and the FPM, we believe that a one-compartment kinetic model is at least as accurate as these other methods. The significant fit achieved between all seven modeled small-bowel TECs and the corresponding measured TECs also attests to the model's accuracy. Furthermore, the additional parameter measurements obtained, and VD, should be considered accurate because the accuracy of any mathematical model depends on the validity of the constants used to define the model. If the , VD, and/or FV values did not accurately reflect small-bowel perfusion, then the modeled TECs that resulted from the insertion of these constants into Equation (1) would not have strongly correlated with the actual measured TECs in every case.
We chose the described kinetic model over other techniques because it involves the use of all portions of the TEC—that is, it takes into account both the inflow and the outflow or washout of small-bowel blood. The FPM and variate analysis were expected to yield FV values similar to the FV values in our model because this parameter is based on only the upsloping portion of the entire bowel TEC and quantifies inflow. However, with the FPM, one assumes that there is no outflow from enhancing tissue, while the variate method does not take into account contrast material recirculation and thus yields underestimated values for the latter portion of the TEC, which reflects washout. In comparison, our kinetic model takes into account the latter portion of the TEC, which is reflected in the VD value. VD is a quantification of the portion of a unit volume of tissue that is available to contrast material; the less available, the faster the washout and vice versa.
The importance of outflow—that is, the clearance of contrast material from tissue—has not been established, but this parameter may be important, especially in bowel tissue, where splanchnic resistance can be altered by a neoplastic process owing to a circulating-tumoral-factor (11) direct mass effect on splanchnic vasculature (12), hepatic disease, or splanchnic thrombosis. We also wish to point out that the regional tissue blood volume (rTBV) determined by using deconvolution techniques assumes that the total amount of blood present within a unit volume of tissue is uniformly distributed; therefore, this value is not the same as VD. VD is a perfusion parameter that cannot be measured with the variate, FPM, or deconvolution techniques.
It should be noted that the shapes of our measured TECs did not appear to indicate that any contrast material washout occurred because there was no decline in small-bowel enhancement during the 120-second period of imaging. However, washout did occur, but it was convoluted owing to contrast material recirculation, with the overall result being steady small-bowel wall enhancement with time. A similar result has been reported for large-bowel wall enhancement (6). If washout had not occurred during the imaging period, Equation (1) would have yielded a value of 0 (Cout[t] = 0), and, thus, Cin(t) and C1(t) would have been linearly related. These findings would have translated into a similar shape for the measured aortic and small-bowel TECs, but these findings were not seen. Our kinetic model allowed the quantification of this washout (represented by VD) even though the washout was convoluted within the TEC.
Similar to VD, —that is, the time it takes for a unit of blood to travel from the arterial inflow source (aorta at the superior mesenteric artery level) to the target tissue—represents another perfusion parameter that is unique to the one-compartment kinetic model. is not similar to the standard mean transit time, which is an estimate of how long blood remains in the target tissue before it exits. We believe that the may also be insightful with respect to alterations in small-bowel perfusion. However, confirming this benefit and that of using VD will require clinical application of our model.
It should be noted that our kinetic model is somewhat complex compared with the FPM and is more difficult to implement than the commercially available CT software based on variate or deconvolution methods. However, the potential usefulness of the additional parameters yielded by our model (VD and ) may offset these limitations.
Another potential limitation of our single-compartment kinetic model is that it does not consider the exchange of contrast material between the intra- and extravascular spaces, and, thus in this aspect, it does not perfectly model tissue perfusion. The application of a two-compartment model to our small-bowel data that takes this exchange into account would be the most optimal. We initially attempted to use such a model but found this to be exponentially more complex and thus probably of little potential clinical use. The fact that the one-compartment model yielded TECs that correlated well with the actual measured TECs indicates that the contrast material exchange between the intra- and extravascular spaces in normal bowel tissue is minimal and negligible. In abnormal small-bowel states, however, this exchange may be enhanced and no longer negligible. Thus, a poor fit between a modeled TEC and an accurately measured TEC may be an indicator of abnormality; however, this theory is speculative and would need to be confirmed.
We also should point out that our model is not unique. Materne et al (13) determined the hepatic perfusion in rabbits by using a one-compartment kinetic model applied to magnetic resonance and CT images. These investigators found that the results from this model significantly correlated with the perfusion measurements obtained by using microsphere contrast agent techniques. A one-compartment kinetic model has also been applied to determine the hepatic perfusion in normal and diseased states (14–16). We still believed it necessary to confirm its competence by comparing it with other more widely used techniques to assess tissue perfusion.
In conclusion, we believe that our study results show that after correction for motion, TECs of the normal small bowel can be constructed and that these curves can be analyzed by using a single-compartment kinetic model to yield perfusion data. This model appears to be as accurate as other models used to assess tissue perfusion; however, by being fit to the entire TEC, this model enables examination of outflow or washout of blood from small bowel in addition to inflow.
|
APPENDIX
|
|---|
The schematic in Figure A1 represents a single-compartment kinetic model as applied to small bowel. Cin(t), C1(t), and Cout(t) represent the concentrations of contrast material at time t within the aorta at the level of the superior mesenteric artery, within the bowel wall, and exiting the bowel wall, respectively. F indicates total blood flow and is constant. For any unit volume of bowel represented by V, V1 represents that portion available to contrast material and V2, the portion unavailable, where V = V1 + V2.
Application of a mass balance to contrast material within bowel indicates the following:
 | (A1) |
where m(t) = contrast mass within V1 at time t and Cin = 0 for 0 < t < . Dividing both sides of Equation (A1) by V and noting that Cout(t) = [m(t)]/V and C1(t) = [m(t)]/V1 yields Equation (1):
|
| (1) |
where FV = F/V and VD = V/V1.
|
FOOTNOTES
|
|---|
Abbreviations: FPM = first-pass method • ROI = region of interest • TEC = time-enhancement curve
Authors stated no financial relationship to disclose.
Author contributions: Guarantors of integrity of entire study, A.S., R.G.S.; study concepts/study design or data acquisition or data analysis/interpretation, A.S., R.G.S.; manuscript drafting or manuscript revision for important intellectual content, A.S., R.G.S.; approval of final version of submitted manuscript, A.S., R.G.S.; literature research, A.S., R.G.S.; clinical studies, A.S., R.G.S.; statistical analysis, A.S.; and manuscript editing, R.G.S.
|
References
|
|---|
- Axel L. Cerebral blood flow determination by rapid-sequence computed tomography: a theoretical analysis. Radiology 1980;137:679–686.[Abstract/Free Full Text]
- Nabavi DG, Cenic A, Craen RA, et al. CT assessment of cerebral perfusion: experimental validation and initial clinical experience. Radiology 1999;213:141–149.[Abstract/Free Full Text]
- Nabavi DG, Cenic A, Henderson S, Gelb AW, Lee TY. Perfusion mapping using computed tomography allows accurate prediction of cerebral infarction in experimental brain ischemia. Stroke 2001;32:175–183.[Abstract/Free Full Text]
- Miles KA. Measurement of tissue perfusion by dynamic computed tomography. Br J Radiol 1991;64:409–412.[Abstract]
- Miles KA, Hayball MP, Dixon AK. Functional images of hepatic perfusion obtained with dynamic CT. Radiology 1993;188:405–411.[Abstract/Free Full Text]
- Kuhle WG, Sheiman RG. Detection of active colonic hemorrhage with use of helical CT: findings in a swine model. Radiology 2003;228:743–752.[Abstract/Free Full Text]
- Spalinger J, Patriquin H, Miron MC, et al. Doppler US in patients with Crohn disease: vessel density in the diseased bowel reflects disease activity. Radiology 2000;217:787–791.[Abstract/Free Full Text]
- Ludwig D, Wiener S, Bruning A, Schwarting K, Jantschek G, Stange EF. Mesenteric blood flow is related to disease activity and the risk of relapse in Crohn's disease: a prospective follow-up study. Am J Gastroenterol 1999;94:2942–2950.[CrossRef][Medline]
- Clough AV, Haworth ST, Hanger CC, et al. Transit time dispersion in the pulmonary arterial tree. J Appl Physiol 1998;85(2):565–574.[Abstract/Free Full Text]
- Marquardt D. An algorithm for least-squares estimation of nonlinear parameters. SIAM J Appl Math 1963;11:431–441.[CrossRef]
- Leen E, Goldberg JA, Anderson JR, et al. Hepatic perfusion changes in patients with liver metastases: comparison with those patients with cirrhosis. Gut 1993;34:554–557.[Abstract/Free Full Text]
- Sheiman RG, Raptopoulos V. Delayed intravenous contrast medium washout from the small bowel in patients with pancreatic carcinoma and splanchnic venous invasion. J Comput Assist Tomogr 1996;20:924–929.[CrossRef][Medline]
- Materne R, Van Beers BE, Smith AM, et al. 2000 Non-invasive quantification of liver perfusion with dynamic computed tomography and a dual-input one-compartment model. Clin Sci 2000;99:517–525.[Medline]
- Van Beers BE, Leconte I, Materne R, Smith AM, Jamart J, Horsmans Y. Hepatic perfusion parameters in chronic liver disease: dynamic CT measurements correlated with disease severity. AJR Am J Roentgenol 2001;176:667–673.[Abstract/Free Full Text]
- Materne R, Smith AM, Peeters F, et al. Assessment of hepatic perfusion parameters with dynamic MRI. Magn Reson Med 2002;47:135–142.[CrossRef][Medline]
- Materne R, Annet L, Dechambre S, et al. Dynamic computed tomography with low- and high-molecular-mass contrast agents to assess microvascular permeability modifications in a model of liver fibrosis. Clin Sci 2002;103:213–216.[Medline]
This article has been cited by other articles:
|
|
|
|
 
S. T. Schindera, R. C. Nelson, D. M. DeLong, T. A. Jaffe, E. M. Merkle, E. K. Paulson, and J. Thomas
Multi-Detector Row CT of the Small Bowel: Peak Enhancement Temporal Window--Initial Experience
Radiology,
May 1, 2007;
243(2):
438 - 444.
[Abstract]
[Full Text]
[PDF]
|
|
|
TOP
ABSTRACT
INTRODUCTION
) obtained from small-bowel ROI analysis of sequential contrast-enhanced images. Lack of decline in enhancement does not indicate absence of contrast washout (which is obscured by the effects of contrast material recirculation).