Computer analysis of radionuclide hepatobiliary studies

zyxw zyxwvutsrqp zyxwvu zyxw zyxwvuts COMPUTER ANALYSIS OF RADIONUCLIDE HEPATOBILIARY STUDIES W. E. Barnes*, J. S. Arnold*#, N. Khedkar*, T. Milo*, E. E. Gose*@ * Nuclear Medicine Service, V . A . Hospital, Hines,,IL d Loyola University, Stritch School of Medicine, Maywood, IL @ Bioengineering Program, University of Illinois, Chicago, IL provided very satisfactorily fits to data from normal patients and cases of common duct obstruction, but proved less satisfactory in cases of cirrhosis.* In order to satisfactorily fit the observed complex changes in the liver uptake curves of cirrhosis, it was necessary to add a second liver compartment. The present paper describes our results using this new compartmental model in ten patient studies with Tc-99m PG. A technique for the computerized analysis of the kinetic behavior of Tc-99m labeled hepatobiliary agents is described. Both blood and extravascular background correction of sequential camera images is used. The results of the analysis of blood, liver and bile concentrations are analyzed using a compartmental model containing two parallel liver compartments. Results in ten patients indicate that in all liver disease cases regurgitation was markedly increased. In complete obstruction of the bile ducts there was irreversible liver uptake, suggesting intrahepatic cholestasis and decreased blood clearance. In alcoholic hepatitis there was an increase in clearance thought to be due to increased hepatic artery blood flow. The present study describes a computerired method of collecting and analyzing sequential images of the liver following the administration of one of the new Tc-99m labeled hepatobiliary agents, pyridoxylideneglutamate (PG). The transfer constants describing the transfer of PG from blood to liver and bile are derived from the fit of the net activity in the blood, liver, and bile as a function of time to a suitable compartmental model having two liver compartments. The transfer constants from blood to liver and liver to bile appear to be very promising, not only as a means of diagnosing the kind of liver disease, but also the severity of the pathophysiologic changes present. Background Corrections Early studies led us to suspect that much of the activity in the liver after injecting a hepatobiliary agent can be accounted for by the activity contained within its rather large blood pool. To eliminate this kinetic problem, we have routinely corrected for the blood contribution by preliminarily administering a small dose of albumin tagged with the radionuclide. The count rate in an area of interest (AOI) over the heart blood pool reflects the activity in the blood throughout the body. In the static albumin blood pool image, the counts in an area of interest such as the liver reflect its blood pool activity alone, while the same area in the dynamic PG study reflects both blood and parenchymal cell concentrated activity. The non-vascular activity in any A01 is given by its gross counts minus the heart counts at that time multiplied by the ratio of the counts in the AD1 in the albumin image to the counts in the heart A01 in the albumin image. We found when using the Tc-99m labeled agents that subtraction of the activity contributed by blood always leaves behind a residual activity forming a diffuse image. This activity is assumed to be present in the extravascular pool of the tissue. Assuming that the contribution of the extravascular background is approximately uniform over the thoracic region, we subtract from all A01 data on a per-pixel basis the blood-subtracted extravascular background as determined over the abdomen or lung where no specific radionuclide concentration occured. The importance of correcting f o r the contribution of blood and extravascular activity is illustrated in Figure 1. The gross counts.over a normal liver are compared to the net counts after subtraction of the blood and of the bloodcorrected counts from an area over the abdomen. The effect of background correction is more dramatic in the case of liver disease when liver uptake is small compared to background. Introduction Curves of the time rate of change of radiotracers in organs have been measured for more than two decades, yielding important physiologic and pathophysiologic concepts. With the advent of rapid digital imaging cameras (Anger type) , the movement of tracer in and out of organs could be recorded and viewed in a cine mode. Using computer analysis of the sequential images, we have moved several steps forward in this process. First, we have developed procedures to strip away the background activity contributed by overlying tissues and by blood so that the curves of the net activity in the specific organ studied can be quantitatively evaluated. Secondly, in the case of the liver, we have constructed a compartmental model which appears to describe the pathophysiologic movement of PG in cirrhosis, hepatitis and obstructive disease of the liver. The model initially contained a single liver compartment and zyxwvutsrq zyxwvu CH1515-6/79/0000-0654$00.75 @ 1979 IEEE 654 zyxwvu zyxwvutsrqp When the work of others in the field of liver kinetics was reviewed, it was noted that no one had carefully corrected for the blood or extravascular background in the liver or for that matter in any of the monitored areas of interest. In only a few studies had data been collected with the Anger camera which provides much more accurate localization of monitored organs. zyx zyxwvutsr zyxw zyxwv zyxwvuts 0 3 0 are recorded: 1) anterior and posterior views of the liver; 2) anterior and posterior views of the entire abdomen; 3 ) the camera injection dose standard. The purpose of the anterior and posterior images of the liver and abdomen is to assess by the method of conjugate views ( 3 ) the total quantity of activity in the liver, gallbladder, and intestine at the end of the kinetic study. The image of the injection dose standard is used to calibrate the images in terms of the administered dose. The liver A01 counts are corrected for the contribution from the gallbladder. The A01 of the intestine is taken as the entire abdominal area not involved with activity from the liver and urinary bladder. All data are corrected for room background and the hepatobiliary agent data are corrected for the contribution from the preliminary injection of Tc-99m labeled albumin. All kinetic data are then corrected for blood and extravascular activity as described above. Two types of compartmental models have been evaluated using data from two normal volunteers and eight patients with various types of liver disease. In the past we have fitted blood, liver and bile data to the compartmental models illustrated in Figures 2 and 3. The two compartment model (Figure 2) has been extensively used in the analysis of 1-131 Rose Bengal ( 4 , 5). The protein binding of 1-131 Rose Bengal produces minimal l o s s from blood to extravascular spaces and urine. However, the limited count rates and relatively slower blood to liver and liver to bile transfer constants than found with the new Tc-99m labeled hepatobiliary agents proved to be serious limitations for clinical use. Iw'o-gross liver 0 .-blood A-body wall o-net liver 0 zyxwv zyxwvutsrqpon 10 20 30 TIME 40 FIGURE 2 . Two compartment model of liver kinetics suitable for 1-131 rose bengal. Tc-99m labeled PG and the iminodiacetic acid (IDA) hepatobiliary agents are not extensively bound to serum proteins and readily pass into the extravascular spaces throughout the body and are excreted in the urine. In order to accomodate extravascular fluid loss and urinary excretion we used a three compartment model (Figure 3) FIGURE 1. Plot of gross liver counts (0) and liver counts (0) corrected for blood (a) and extravascular tissue (A). Methods Our approach has been to place the patient to be studied in a supine position under a large field of view Anger camera so that the heart, liver and abdomen are viewed anteriorly. Two hundred uCi of Tc-99m labeled human albumin is administered I.V. and images and counts collected from 5 to 10 minutes after injection. Two millicuries of Tc-99m PG are then administered I.V. and 30 second images are sequentially recorded for up to 60 minutes in 96 x 96 matrices and stored on a magnetic disc. Areas of interest are placed over a peripheral region of the liver (remote from the gallbladder and major bile ducts), the gallbladder, heart, intestine, and lung or abdominal background areas. The A01 data from all areas are read into a Nova 2 computer. Immediately following the completion of the above sequential study the following static views 1 URINE FIGURE 3. 655 Three compartment model incorporating urinary excretion and loss to the extravascular space. Regurgitation from liver to blood and bile excretion from the liver are measured together as the loss of activity from the liver in the following way: Liver activity at any time t may be expressed as zyxwvutsrqpon zyxwv zyxwvutsr Alternatively, the first liver compartment could excrete to bile whi.le also communicating with a second liver compartment: where L = activity in the liver B = activity in the blood i 1 1 BLOOI 1-r KBL = fraction of activity in the blood transferred to liver per unit time KLB = fraction of activity in the liver transferred to blood per unit time K = fraction of activity in the liver transferred to bile per unit time BG t.c As we visualize hepatic disease, it is very nonuniform at the tissue level. There is a mixture of normal lnbulesx as well as abnormal lobules. Both types of tissue communicate with blood independently, and do not communicate with each other. On this basis we have chosen a model with parallel liver compartments independently communicating with blood: By numerically integrating the blood and liver activity curves over time, it is straightforward to determine KBL and (KLB f by linear least squares fitting. It should be noted that it is not necessary to know the transfer constant from blood to urine o r the transfer constants to and from the extravascular compartment to make these determinations. Although loss to urine and to the extravascular compartment alter the shape of the blood curve, they do not affect the transfer constants between blood, liver and bile. In order to separately measure regurgitation and bile excretion, we devised a technique for measuring the rate of excretion into bile, knowing the activity ( G ) present in the gallbladder and intestine at the end of the 50 minute kinetic study. zyx zyxwvutsrqpo G = KLG fL In fitting data to this model, we found that satisfactory fits could only be made when the compartment excreti-ng to bile (Liver #l) was that compartment having the fastest turnover time. In fact, satisfactory fits could be obtained without any route of egress from the second liver compartmen t Determination of the transfer constants between blood and the two liver compartments proceeds in much the same way as with the single liver compartment except that a non-linear least squares fitting routine is used. We have found the iterative method of Marquardt (7) quite satisfactory. An initial estimate of the transfer constant values is made by assuming that at the conclusion of the study a l l the activity resides in the slow-turnover compartment and calculating the transfer constants of the slow compartment by linear least squares fitting.lhis component is then subtracted from the liver curve and the remainder fit with the fast-turnover compartment. . dt The value of the regurgitation constant from liver back to blood can then be calculated as: zyxwvutsrqpon The data from normal patients and those with complete biliary obstruction (such as is caused by cancer of the bile ducts) can be fitted well to the three compartment model (2, 6 ) . However, the fit of the data from alcoholic hepatitis patients was defective in that the model could not adequately accomodate the early spike in liver activity occuring at about five minutes after injection as illustrated in Figure 4 B . Clearly, a second liver compartment is required to accomodate this pathophysiologic change. In considering the possible arrangements of a second liver compartment, several alternatives exist. One possible arrangement has the two liver compartments in series: Results 'I%e results from the analysis of Tc-99m PG patient studies are presented in two forms: 1) The curves of background corrected blood, liver and bile; 2) the calculated transfer constants between blood, liver and bile. * The liver lobule is a basic structural unit composed of liver cells, blood vessels and bile ducts. 656 zyxwvutsrqp zyxwv zyxwvut In Figure 4 the characteristic net liver curves representative of (A) normals, (B) alcoholic hepatitis, and (C) complete biliary obstruction patients are presented. It is apparent that in normals there is a prompt rise in liver activity followed by a slow progressive loss of activity (into bile). In alcoholic hepatitis (Figure 4B), however, there is a faster rise in liver activity immediately following injection than seen in normals. Once acquired in the liver, the activity is rapidly lost again starting after five minutes producing a spike-like elevation in the curve. In obstructive biliary disease (Figure 4C) there is a slower than normal rise in the liver curve which continues throughout the observation period. c3 4 through the hepatic sinusoids. The extraction efficiency is thought to be a function of the availability of binding sites on the cell membrane as well as the strength of binding of these sites for the radiopharmaceutical. The more rapid rise in liver activity in alcoholic hepatitis may represent an increase in either the strength of binding sites available on hepatic cells or to an increase in the rate of blood flow through the liver. The blood flow to the liver is known to be of two types: 1) arterial blood from the hepatic artery and 2) venous blood from the portal vein. In cirrhosis (8) and alcoholic hepatitis (9) the hepatic artery blood flow is known to be increased while the portal vein blood flow is decreased. Because the portal vein blood must first perfuse the intestine or spleen on its way to the liver there is a 15 to 30 second lag time in the arrival of activity delivered by portal vein blood as compared to thar delivered by the hepatic artery. The rapid rise of the liver curve in alcoholic hepatitis could partially reflect the greater hepatic artery blood flow and its slightly earlier arrival than the portal vein blood seen in normals. The rapid decrease in activity on the down-side of the spike in alcoholic hepatitis suggests rapid loss of the bound activity back to blood which is called regurgitation. In the case of the complete biliary obstruction cases, the slow progressive rise in liver curves reflects two processes. The slow initial rise suggests less net clearance of the agent from the blood than normal. Since blood flow is probably normal this would suggest a decrease in hepatic blood clearance. Since blood levels in these cases were slowly falling due to excretion of activity in the urine some active process had to be progressively concentrating activity in the liver. Since no activity was excreted into bile ducts it is assumed that activity was being concentrated in the dilated bile canaliculi characteristic of obstructive biliary disease (bile stasis or cholestatis). This may be responsible for the slow rise in the late portions of liver curves of some of the cirrhotic patients with alcoholic hepatitis. zyxwvutsrqponmlkjihg w zyxwvutsrqp zyxwvutsrqpo ALCOHOLIC 20 P H E PA1I1 I S U U I O 1 20 zyxwvutsrqpon m m 1 30 4 0 5 0 MIN 10 30 20 MIN 40 Transfer constants in disease states The transfer constants which resulted from the best fit of the data from ten patients to the compartmental model (with two parallel liver compartments communicating with blood) are presented in Table 1. Cases 1-3 are normal cases or not thought to have any liver disease. Case 4 is a case of chronic progressive hepatitis. Cases 5-7 are cases of acute alcoholic hepatitis who also had cirrhosis of the liver. Case 8 is a case of partial obstruction of the bile duct due to a gallstone and may also have had some component of cirrhosis. Cases 9 and 10 are both carcinomas obstructing the common bile duct for several weeks prior to study. Inspection of Table 1 reveals that the values for K1, which are measures of hepatic blood clearance, are quite high in two of the three alcoholic hepatitis cases suggesting an increased blood flow or extraction efficiency. The value of K1 was zyxwvutsrq zyx 50 FIGURE 4. Liver curves representative of normal subjects ( A ) , alcoholic hepatitis (B), and biliary obstruction (C). Physiologic Interpretation of Curves The initial rise in the liver curves must be correlated with a) the rate of blood flow to the liver and/or b) the extraction efficiency of the hepatobiliary agent from the blood as it passes 657 zyxwvutsrqpo zyxwvutsrqp zyxwvutsrq zyxwvuts zyxwvutsrqp zyxwvutsr normal or low in the two cases of complete bile duct obstruction. Table 1. Transfer constants between blood, liver, and bile expressed as percent of activity transferred per minute. y5 -- Normal 6 Gastric ulcer Gastri t1s Chronic hepatitis Alcoholic hepatitis Alcoholic hepatitis Alcoholic hepatitis Partial obstruction Complete obstruction Complete obstruction 14 5 3 3 2 9 74 22 20 10 6 5 2 K3 6.2 1.7 7.1 5.1 1.8 K4 0 0 0 0 0 0 0.3 3.1 to indicate there was little or no bile excretion. The liver curve in this patient was the same shape as the blood curve indicating that reversable retention at binding sites in the hepatic cells was the principal location of liver activity. The value which seems to best separate the obstructive cases from alcoholic hepatitis cases is the percent that cholestatic retention represents of the total blood clearance of the liver (K4/K X 100). In alcoholic hepatitis it is 0 to 8.48 while it is 17-48% in the two cases of complete bile duct obstruction. Conclusion The availability of rapid digital radioisotope imaging and computer analysis makes it possible to not only see to what organs an injected radiopharmaceutical is going, but to measure the rates. Once the rates of movement have been measured, the practical question arises as to whether a difference in rate exists between normals and patients with disease. Lastly, can the rates measured in one disease be differentiated from the rates measured in another kind of disease? At this point in our research it is apparent that we are able to detect disease and, based on a small series of studies, to differentiate between different diseases. Large groups of cases with known diagnoses will have to be analyzed to evaluate the diagnostic effectiveness of the procedure. The most important contribution of our work has been to show that the most significant kinetic information in hepatobiliary studies occurs in the first fifteen minutes after injection when blood flow and regurgitation are defined. During this period blood background is very high, swamping the liver uptake information. Unless blood and body wall background are subtracted, there is no possibility of accessing blood clearance and regurgitation information. zyxwvutsrq 28 50 32 25 26 16 2.8 5.4 4.5 1.9 0.8 0 0.6 0.8 0.4 0.8 1.1 0 2.8 8.4 6.6 16.7 47.8 The values of K2 are thought to represent loss of activity from the binding sites of the liver-cells as a consequence of the cells not being able to transport the concentrated material across the cell to be excreted into the bile canaliculi. Its values are highest in the single case of chronic hepatitis where 74% of the activity in the liver is regurgitated back to blood per minute. F r o m 28-32% regurgitation per minute occurs in the three cases of alcoholic hepatitis. In biliary obstruction only slightly less regurgitation is present than seen in alcoholic hepatitis. Kg is the X of liver activity which is excreted into the bile per minute. This value is seen to vary from 7 to 1%. In the two cases of complete biliary obstruction the value should be zero and therefore the values of 1.9 and .8 represent the error due to inadvertently picking abdominal background areas of interest which were lower than the average background. It is difficult to attribute any significance to the values in the non-obstructed cases since they appear to vary randomly at comparable levels in all groups. The values of K4 represent activity which was irreversibly bound in the liver, producing the slowly rising component in the late liver curves of biliary obstruction and alcoholic hepatitis. The values probably represent the process of intrahepatic cholestasis, expressed in units of percent of the blood activity transferred per minute to the static liver compartment. K4 is noted to be slightly higher in the cases of complete bile duct obstruction than in alcoholic hepatitis. One case of alcoholic cirrhosis shows a zero value for IC4 indicating that no cholestatic retention was present and the value of K3 was sufficiently low zyxwvutsrqpon zyxwvutsrqp References 1. Arnold J S , Barnes WE, Shponka S, et al: A universal approach to tissue and blood backaround correction of static and dvnamic camera images. In Proceedings of-Sixth Symposium on Sharing of Computer Programs and Technology in Nuclear Medicine, Society of Nuclear Medicine, 1976, pp 361-369. 2. Arnold.JS, Barnes WE, Khedkar N, Nelson M: Computer analysis of hepatobiliary studies in the diagnosis of liver disease. J Nucl Med 20, 686, 1979. 3. Budinger TF: Quantitative nuclear medicine imaging: application of computers to the gamma camera and whole-body scanner. In Recent Advances in Nuclear Medicine, ed Lawrence JH, vol 4, pp 41-130, 1974. 4. Turco GL, Gehmi F, Molino G, et al: The kinetics of 1-131 rose bengal in normal and cirrhotic subiects studied bv comvartmental analysis and digital compuier. j Lab Clin Med 7, 983, 1966. - a 5. Waxman AD, Leins PA, Siemsen JK: In vivo dynamic studies of hepatocyte function: A computer method for the interpretation of rose bengal kinetics. Comput Biomed Res 5, 1, 1972. 6. zyxwvutsrqp zyxwvutsrq zyxwvutsr zyxwvutsr zyxwvutsrqpo Arnold JS, Barnes WE: Dynamic studies with hepatobiliary agents. In Proceedings of Seventh Symposium on Sharing of Computer Programs and Technology in Nuclear Medicine, Society of Nuclear Medicine, 1977, pp 254-267. 7. Marquardt DW: A n algorithm for least-squares estimation of nonlinear parameters. J SOC Ind Appl Math 11, 431, 1963. 8. DeNardo GL, Stadalnik RC, et al: Hepatic scintiangiographic patterns. Radio1 111, 135, 1974. 9. Cohn JN, Khatri MB, et al: Hepatic blood flow in alcoholic liver disease measured by an indicator dilution technic. Am J Med 53, 704, 1972. - 659