Hepatobiliary Imaging Original Research. rior researchers have described. the usefulness of biphasic contrast-enhanced

Transcription

1 MDCT of the Liver and Hypervascular Hepatocellular Carcinomas Hepatobiliary Imaging Original Research A C D E M N E U T R Y L I A M C A I G O F I N G MDCT of the Liver and Hypervascular Hepatocellular Carcinomas: Optimizing Scan Delays for Bolus-Tracking Techniques of Hepatic Arterial and Portal Venous Phases Satoshi Goshima 1 Masayuki Kanematsu 1 Hiroshi Kondo 1 Ryujiro Yokoyama 2 Toshiharu Miyoshi 2 Hironori Nishibori 1 Hiroki Kato 1 Hiroaki Hoshi 1 Minoru Onozuka 3 Noriyuki Moriyama 4 Goshima S, Kanematsu M, Kondo H, et al. Keywords: contrast media, CT technique, liver DOI:1.2214/AJR Received December 13, 24; accepted after revision May 9, Department of Radiology, Gifu University School of Medicine, 1-1 Yanagido, Gifu , Japan. Address correspondence to S. Goshima (gossy@par.odn.ne.jp). 2 Department of Radiology Services, Gifu University School of Medicine, Gifu, Japan. 3 Department of Physiology and Neuroscience, Kanagawa Dental College, Yokosuka, Japan. 4 Department of Diagnostic Radiology, National Cancer Center Hospital, Tokyo, Japan. WEB This is a Web exclusive article. AJR 26; 187:W25 W X/6/1871 W25 American Roentgen Ray Society OBJECTIVE. The purpose of our study was to determine the optimal scan delays required for hepatic arterial and portal venous phase imaging and for the detection of hypervascular hepatocellular carcinomas (HCCs) in contrast-enhanced MDCT of the liver using a bolus-tracking program. SUBJECTS AND METHODS. CT images (2.5-mm collimation, 5-mm thickness with no intersectional gap) detected an increase in the CT value of 5 H in the lower thoracic aorta. The images were obtained after an IV bolus injection of 2 ml/kg of nonionic iodine contrast material (3 mg I/mL) at 4 ml/s in 171 patients, who were prospectively randomized into three groups with scans commencing at 5, 2, and 45 seconds; 1, 25, and 5 seconds; and 15, 3, and 55 seconds for the first (acquisition time: 4.3 seconds), second (4.3 seconds), and third (9.1 seconds) phases, respectively, after a bolus-tracking program. CT values of the aorta, spleen, proximal portal veins, liver parenchyma, and hepatic veins were measured. Increases in CT values from unenhanced to contrast-enhanced CT were assessed using a contrast enhancement index (CEI). Spleen-to-liver and HCC-to-liver contrasts were also assessed. A qualitative degree of contrast enhancement in each organ was prospectively assessed by two independent radiologists. RESULTS. At 1 15 seconds, the CEI of the aorta reached H and that of the spleen reached H without significant enhancement of liver parenchyma (15 25 H). The CEI of the proximal portal veins moderately increased (75 14 H) at 1 15 seconds, but no significant enhancement of hepatic veins was observed (24 51 H). The CEI of liver parenchyma peaked (59 63 H) at seconds, when the CEIs of the aorta ( H) and spleen (73 82 H) decreased. Spleen-to-liver contrast (81 84 H) was highest at 1 2 seconds and HCC-to-liver contrast (39 44 H) was highest at 1 15 seconds. The qualitative results correlated well with quantitative results. CONCLUSION. The optimal scan delays for hepatic arterial and portal venous phases after the bolus-tracking program detected threshold enhancement by 5 H in the lower thoracic aorta for the detection of hypervascular HCCs were 1 15 and seconds, respectively. rior researchers have described P the usefulness of biphasic contrast-enhanced CT of the liver for detecting hypervascular malignant hepatic tumors, typically hepatocellular carcinomas (HCCs) [1, 2] or hypervascular metastases [3]. The ability of contrast-enhanced CT to detect hepatic tumors is affected by several radiologic factors, such as dose [4, 5], concentration [6], injection rate [7 9] of iodine contrast material, and scan delay after contrast injection [1], and by pathologic factors such as size, histologic grade, or tumor vascularity [11]. Several researchers have investigated time density profiles for various doses, concentrations, and injection rates of contrast material using single-detector helical CT. Foley et al. [8] reported that hepatic enhancement increased to more than 5 H 6 seconds after initiation of monophasic IV injection of contrast material with a 5-g iodinated load at 3 ml/s. Tublin et al. [9] reported that peak enhancement of the liver occurred 63 and 87 seconds after initiating contrast material injection of 15 ml of 3 mg I/mL at rates of 5. and 2.5 ml/s, respectively. Also, Lee et al. [12] reported that the optimal duration for the hepatic arterial dominant phase was from 36 to 56 seconds after the injection of 15 ml of contrast material at a rate of 3 ml/s, but added that no optimal fixed delay time is appropriate for all patients. These previous studies evaluated the time density profiles from the initia- AJR:187, July 26 W25

2 tion of contrast medium injection without incorporating circulation time difference in individual patients, which significantly influence peak enhancement time [7, 13, 14]. With the advent of MDCT, the acquisition time for whole liver imaging is fewer than 5 1 seconds. Such a short acquisition time provides an opportunity to capture peak enhancement in visceral organs by optimizing contrast medium injection and scan protocols. However, an off-timing scan markedly increases the risk of suboptimal contrast enhancement. A bolus-tracking technique has become widely available and may be indispensable for the optimization of scan timing in individual patients to compensate for the variability of circulation time between patients. Although several researchers have investigated the importance of bolus-tracking [15 17], no previous reports on the optimization of scan delay for the bolus-tracking technique have, to our knowledge, been available. Therefore, the purpose of our study was to optimize scan delays for hepatic arterial and portal venous phases for bolus-tracking techniques in MDCT of the liver. Subjects and Methods Patients During a 6-month period (June November 22), 29 consecutive patients suspected of having hepatic disease, and who had previously undergone sonography, MRI, or laboratory evaluation, underwent contrast-enhanced CT of the upper abdomen at our department. All patients were informed that the radiologic examinations were primarily for clinical diagnosis and secondarily for radiologic research. Thereafter, they all provided written consent in accordance with the requirements of our institutional review board. The study was performed according to the guidelines of the Declaration of Helsinki [18]. Patients were prospectively randomized into three imaging groups. We excluded 38 patients from our study for the following reasons: five patients had numerous recurrent tumors after treatment, 15 patients had undergone partial hepatectomy for malignant hepatic tumors, six had undergone total splenectomy, four had diffuse or segmented fatty liver of a severe degree, and eight experienced technical failure related to contrast medium injection, breath-holding, or a machine problem during CT examination. These patients were excluded from the study population because hepatic vascular alterations caused by numerous tumors or preceding surgery, abnormal hepatic density, or artifacts might have adversely affected our evaluation of contrast enhancement. The remaining 171 patients comprising 12 men and 69 women (age range, years; mean, 63.1 years) constituted the study population. The study population contained 27 patients with HCC in cirrhosis, three with HCC in noncirrhotic livers, 17 with liver metastases (from colorectal [n = 8], gastric [n = 4], pulmonary [n = 3], duodenal [n =1], and uterine cervical [n = 1] cancers), eight with chronic hepatitis, three with cavernous hemangiomas, and 113 with extrahepatic primary neoplasms and healthy livers (gastric [n = 27], colorectal [n = 26], lung [n = 2], uterine [n = 12], lymphoma [n = 5], esophageal [n =4], ovarian [n = 3], malignant melanoma [n = 2], cholangiocarcinoma [n = 2], breast [n =2], renal [n = 2], pancreatic [n = 2], laryngeal [n = l], cholelithiasis [n = 1], acute pancreatitis [n = 1], chronic pancreatitis [n = 1], urinary stone [n = 1], and an unknown fever [n =1]). Contrast Material Injection and Scan Protocols A CT scanner (LightSpeed Ultra, GE Healthcare) with an mm detector configuration was used, in which eight interspaced helical data sets were collected from eight detector rows. The high-speed mode used for first- and second-phase CT in our study was equivalent to a helical pitch of 1:1.35 with a table speed set at 27 mm per rotation (.5 seconds). The high-quality mode used for the third-phase CT was equivalent to a pitch of 1:.875 with a table speed set at 17.5 mm per rotation (.7 seconds). These scanning parameters were selected to scan the entire liver as rapidly as possible without impairing image quality. Transverse images were reconstructed and displayed as 4 sections of 5-mm-thick images with no intersectional gap for each phase set. All patients were administered nonionic iodine contrast material containing 3 mg I/mL (Omnipaque 3, Daiichi Pharmaceutical) using a power injector (Autoenhance A-5, Nemotokyorindo) at a Group 1 Group 2 Group 3 1st 1st 1st 2nd 2nd rate of 4 ml/s through a 21-gauge plastic IV catheter placed in an upper extremity vein, typically in an antecubital vein. The volume of contrast material delivered was 2 ml/kg of body weight in patients ranging in weight between kg (n = 165), but was fixed at 15 ml for patients weighing between kg (n = 6), which resulted in a total volume of contrast material of ml (mean, 112 ml). A bolus-tracking program (Smart Prep, GE Healthcare) was used to commence the diagnostic scans after contrast injection. This enabled real-time monitoring and the automatic calculation of CT values in a region of interest, and manual initiation of a diagnostic scan after the CT value reached a threshold. The region-of-interest cursor for bolus tracking was placed in the aorta at a level just above the diaphragmatic dome; this level was also used as a starting position for diagnostic scans. Real-time lowdose (12 kvp, 5 ma) serial monitoring scans were initiated 5 seconds after the start of contrast medium injection. During the 5-second interval between contrast injection start and the monitoring scan start, patients were carefully observed by a radiologist for extravasation or acute adverse events caused by the contrast medium injection. Patients were randomized into three groups so that scans were commenced at 5, 2, and 45 seconds; 1, 25, and 5 seconds; and 15, 3, and 55 seconds for the first (acquisition time, 4.3 seconds), second (acquisition time, 4.3 seconds), and third (acquisition time, 9.1 seconds) phases, respectively, after the trigger. This trigger threshold level was set to increase the CT value by 5 H at the lower thoracic aorta just above the level of the diaphragmatic dome. The aortic transit time of contrast material was also calculated as the time from initiation of contrast material administration to the trigger point determined by the bolus-tracking program. Figure 1 illustrates the timing schemes used 3rd Breathing Interval Fig. 1 Diagram illustrating timing scheme. Patients were prospectively randomized into three groups, and scans were initiated at 5, 2, and 45 seconds; 1, 25, and 5 seconds; and 15, 3, and 55 seconds for first (acquisition time, 4.3 seconds), second (acquisition time, 4.3 seconds), and third (acquisition time, 9.1 seconds) phases, respectively, after bolus-tracking program detected increase in CT value of 5 H in aorta just above level of diaphragmatic dome. First- and second-phase images were obtained during single breath-hold, followed by 2-second breathing interval before third-phase scan. Equilibrium phase scan commenced 16 seconds after bolus-tracking trigger. 2nd 3rd 3rd W26 AJR:187, July 26

3 MDCT of the Liver and Hypervascular Hepatocellular Carcinomas for the three imaging protocols. The first- and second-phase scans were completed during a single breath-hold, and the third-phase scan was commenced after a 2-second breathing interval. The equilibrium phase scan was commenced 16 seconds after the trigger in all patients, although these images were not evaluated in the current study. Quantitative Image Analysis Mean CT values (H) in the aorta, spleen, right and left proximal portal veins, liver parenchyma, and hepatic veins were measured in all patients on the CT console monitor by using a circular region-of-interest cursor ranging in size from 5 to 3 mm in diameter on the unenhanced, first-, second-, and thirdphase images. CT values in the aorta were measured in areas just above the level of diaphragmatic dome; in the spleen, they were measured in one area covering as much area of splenic parenchyma as possible; in the proximal portal veins, values were measured in two areas (in the right and left proximal portal branches) and then averaged; in the liver parenchyma, they were measured in three areas (the right anterior segment, the right posterior segment, and the left lobe of the liver) and then averaged; and in the hepatic veins, they were measured in three areas (the right, middle, and left hepatic veins) and then averaged. Focal hepatic or splenic lesions, blood vessels, bile ducts, calcifications, and artifacts were carefully excluded from all measurement areas. We performed a subanalysis of 3 patients who had 36 untreated HCCs ranging from 5 to 75 mm (mean, 16.4 mm). The 3 patients included 21 men and 9 women ranging in age from 54 to 92 years (mean, 69 years) who had chronic liver damage caused by type B (n = 5) or type C (n = 22) hepatitis or with no underlying liver disease (n = 3). Proof of the untreated HCC was obtained with definitive surgery (n = 6), percutaneous liver biopsy (n =12), or measurement of substantially increased α-fetoprotein or proteins induced by vitamin K antagonism (PIVKA-II) levels, with follow-up CT or MRI showing the progression of hepatic tumors (n = 12). CT values were measured on unenhanced, first-, second-, and third-phase images. A circular cursor was placed to encompass as much of the HCC as possible. When the lesion was too small to accurately place a cursor, the image was magnified up to three times. Quantitative degrees of contrast enhancement were expressed as contrast enhancement indexes (CEIs), which were calculated by subtracting CT values on unenhanced images from those on contrast-enhanced images. Spleen-to-liver contrast was assessed as a CT value difference, by subtracting the CT value of the liver from that of the spleen. HCCto-liver contrast was similarly assessed by subtracting a CT value of the liver from that of the HCC. Qualitative Image Analysis Two independent gastrointestinal radiologists prospectively reviewed the first-, second-, and third-phase images separately versus unenhanced images. Images were evaluated subjectively by the two interpreters, who were blinded to patient clinical information. Each interpreter evaluated images in terms of the degree of contrast enhancement in anatomic structures: the proper hepatic artery in the porta hepatis, spleen, proximal portal veins, liver parenchyma, and hepatic veins. Interpreters used a four-point scale: a score of was assigned when an organ showed virtually no enhancement, 1 for minimal to mild enhancement, 2 for moderate enhancement, and 3 when an organ was maximally enhanced. The degree of splenic enhancement moiré pattern was also assessed, using a four-point scale: a score of for no moiré pattern, 1 for a mild pattern, 2 for a moderate pattern, and 3 for a prominent pattern. A splenic enhancement moiré pattern was defined as a heterogeneous enhancement of the spleen, which is believed to be caused by its unique anatomic structure, with variable rates of flow through the cords of red and white pulp. Each interpreter also evaluated HCC depiction for the hepatic arterial enhancement and washout degree. Interpreters used a four-point scale: a score of was assigned when a HCC showed no arterial enhancement or washout, 1 for minimal to mild enhancement, 2 for moderate enhancement, and 3 for marked arterial enhancement or clear washout. When a disagreement occurred, consensus was reached by discussion. Statistical Analysis Analysis of variance and multiple comparisons with the Scheffé criterion [19] were used to evaluate the following determinants in the three groups; patient age; patient body weight; aortic transit time; spleen-to-liver contrast; and CEIs of the aorta, spleen, proximal portal veins, liver parenchyma, and hepatic veins. The Kruskal-Wallis test and multiple comparisons with the Scheffé criterion were used to evaluate qualitative scores obtained as categoric data. To assess interobserver variability at interpreting images, kappa statistics were used to measure the degree of agreement. A kappa value of up to.2 stood for slight agreement,.21.4 for fair agreement,.41.6 for moderate agreement,.61.8 for substantial agreement, and.81 or greater for almost perfect agreement. Results The three patient groups each contained 57 patients, with no significant differences in patient age, patient body weight, aortic transit time, and number of patients with chronic liver damage among the three groups (Table 1). The mean CEIs of the aorta, spleen, proximal portal veins, liver parenchyma, and hepatic veins in the three patient groups are summarized in Table 2. The scan delays after the trigger-versus-mean CEI curves for the aorta are shown in Figure 2; those for the spleen, proximal portal veins, liver parenchyma, and hepatic veins are shown in Figure 3. The mean CEI of the aorta showed a peak (336 H, Fig. 2) at a scan delay of 1 seconds after the trigger (density of the aorta reaching 5 H when obtained at the level of the diaphragm), and then was progressively less with time (Fig. 2). The mean CEI of the spleen increased constantly at images obtained with a scan delay of 5 2 seconds, peaked ( H) at a scan delay of 2 25 seconds, and then gradually reduced with a scan delay of 3 to 55 seconds (16 73 H) (Fig. 3). The mean CEI of proximal portal veins increased constantly at images obtained with a scan delay of 5 to 25 seconds, peaked (149 H) at a scan delay of 25 seconds, and then gradually reduced (Fig. 3). The mean CEI of liver parenchyma increased gradually at images obtained with a scan delay of 5 to 3 seconds, exceeded 5 H at a scan delay of 3 sec- TABLE 1: Patient Age, Body Weight, and Aortic Transit Time in the Three Groups Scan Delay Group 1-, 25-, 5-s Delays 15-, 3-, 55-s Delays Factor Imaging Delay Group (n = 57) (n = 57) Age (y) 18 8 (62.9 ± 11.8) b (64. ± 12.3) b (62.1 ± 13.5) b Body weight (kg) 38 8 (55.8 ± 9.5) b (56.8 ± 9.9) b (57.8 ± 11.4) b Aortic transit time (s) a 6 25 (14.5 ± 3.8) b 9 22 (14.5 ± 2.6) b 6 33 (15.4 ± 5.1) b No. (%) of patients with 14 (25) 13 (23) 8 (14) chronic liver damage Note No significant differences were seen between groups for any factor. a Time from initiation of contrast administration to trigger point determined by bolus-tracking program. b Data in parentheses are mean ± 1 SD. AJR:187, July 26 W27

4 TABLE 2: Results of Quantitative Measurements in the Three Groups First-Phase Scan Delays Second-Phase Scan Delays Third-Phase Scan Delays Anatomic Structure 5 s 1 s 15 s 2 s 25 s 3 s 45 s 5 s 55 s Abdominal aorta ± ± ± ± ± ± ± ± ± 21.9 Spleen 66.6 ± ± ± ± ± ± ± ± ± 13.6 Main portal veins 42.5 ± ± ± ± ± ± ± ± ± 19.7 Hepatic parenchyma 1.7 ± ± ± ± ± ± ± ± ± 46.1 Hepatic veins 19.8 ± ± ± ± ± ± ± ± ± 16.9 Note Data are mean ± 1 SD. Index (H) Fig. 2 Graph showing scan delay after trigger-versus-mean CEI for aorta. Scan delay is amount of time after bolustracking program detected threshold enhancement of 5 H in aorta. Mean CEI of aorta in first phase showed peak at 1 seconds after trigger and then began to decrease constantly with time. Mean CEI of aorta was significantly higher at 1 seconds than at 15 seconds (p <.1) and higher at 2 seconds than at 25 3 seconds (p <.1), respectively. Note that x-axis is indicative not of time course (repeated measurement) but of three different subgroups with different imaging delays, comprising 57 patients each. Error bars = standard errors of means. Index (H) Fig. 3 Graph showing scan delay after trigger-versus-mean CEI for spleen, proximal portal veins, liver parenchyma, and hepatic veins. Mean CEI of spleen showed peak at 2 seconds. Mean CEI of spleen was significantly higher at 1 15 seconds than at 5 seconds (p <.1) and higher at 2 seconds than at 3 seconds (p <.5). Mean CEI of proximal portal veins increased constantly from 5 to 25 seconds, and had peak at 25 seconds. Mean CEI of proximal portal veins was significantly higher at 15 seconds than at 5 1 seconds (p <.1) and higher at 25 3 seconds than at 2 seconds (p <.5). Mean CEI of liver parenchyma increased constantly from 5 to 3 seconds and then plateaued at seconds. Mean CEI of hepatic veins peaked at 45 seconds. Note that x-axis is indicative not of time course (repeated measurement) but of three different subgroups with different imaging delays, comprising 57 patients each. Error bars = standard errors of means. onds (52 H), and then plateaued (59 63 H) at a scan delay of seconds. The mean CEI of the hepatic veins increased constantly at images 5 1 Spleen Proximal portal veins 8 Liver parenchyma 6 Hepatic veins 4 obtained with a scan delay of 2 to 3 seconds and then plateaued ( H) at a scan delay of seconds (Fig. 3). The scan delays after the trigger-versusmean spleen-to-liver contrast curves are shown in Figure 4. Spleen-to-liver contrast had peaked (81 84 H) at a scan delay of 1 2 seconds and decreased from a scan delay of 25 seconds. The scan delays after the trigger-versus-mean HCC-to-liver contrast curves and the mean CEI of the liver with chronic liver damage and HCCs in the 3 patients are shown in Figure 5. The HCC-toliver contrast peaked (39 44 H) at a scan delay of 1 15 seconds, decreased from a scan delay of 25 seconds, and was below zero at a scan delay of seconds. The mean CEI of liver parenchyma in the 3 patients with HCCs increased gradually at images obtained with a scan delay of 5 to 3 seconds, exceeded 5 H at a scan delay of 3 seconds (5 H), and then plateaued (51 59 H) at a scan delay of seconds. The scan delays after the trigger-versusmean qualitative contrast enhancement relations are shown in Figure 6. The mean degree of proper hepatic arterial enhancement was consistently high at a scan delay of 5 15 seconds. The mean degree of splenic enhancement peaked at a scan delay of 2 seconds and then reduced with time. The moiré pattern splenic enhancement was high at a scan delay of 5 15 seconds, but no moiré pattern was observed at a scan delay of seconds. The mean degree of proximal portal venous enhancement increased at images obtained with a scan delay of 5 25 seconds and plateaued at a scan delay of seconds. The mean degree of liver parenchymal enhancement constantly increased at images obtained with a scan delay of 5 to 45 seconds and peaked at a scan delay of seconds. Qualitative results corresponded well with quantitative results (Figs. 2 5). The scan delays after the trigger-versus-qualitative tumor arterial enhancement, washout curves, and mean degree of liver with chronic liver damage and HCCs are shown in Figure 7. The mean degree of tumor arterial enhancement W28 AJR:187, July 26

5 MDCT of the Liver and Hypervascular Hepatocellular Carcinomas was highest at a scan delay of 1 2 seconds (p <.1) and decreased with time. The mean degree of tumor washout was high at a scan delay of seconds and peaked at a scan delay of 55 seconds. The mean degree of liver parenchymal enhancement in the 3 patients with HCCs constantly increased at images obtained with a scan delay of 5 to 45 seconds and then plateaued. Qualitative results corresponded well with quantitative results (Figs. 2 5). The kappa values for the two interpreters ranged from.78 to.91 (mean,.86) for rating the images independently, indicating substantial to almost perfect agreement. Discussion Most HCCs exhibit hyperattenuation against the background liver parenchyma during the hepatic artery phase of contrastenhanced CT, and the optimal scan delay for the hepatic artery phase is crucial for their detection [1 4, 11, 2 26]. Bae et al. [14] stud- Spleen-to-Liver Contrast (H) 1 5 Index (H) Liver with chronic liver damage and HCC HCC-to-liver contrast Fig. 4 Graph showing scan delay after trigger-versus-mean spleen-to-liver contrast curve. Spleen-to-liver contrast in first phase had peak at 1 to 2 seconds and began to reduce at 25 seconds. Mean spleen-to-liver contrast was higher at 1 15 seconds than at 5 seconds (p <.1). Note that x-axis is indicative not of time course (repeated measurement) but of three different subgroups with different imaging delays, comprising 57 patients each. Error bars = standard errors of means. Fig. 5 Graph showing scan delay after trigger-versus-mean hepatocellular carcinoma-to-liver (HCC-to-liver) contrast and CEI of liver with chronic liver damage and HCCs. Mean HCC-to-liver contrast was high (39 44 H) at 1 15 seconds, gradually reduced, and then fell below zero at seconds. But there was no significant difference in HCC-to-liver contrast. Mean CEI of liver parenchyma in patients with HCCs increased constantly from 5 to 3 seconds and then plateaued at seconds. Note that x-axis is indicative not of time course (repeated measurement) but of three different subgroups with different imaging delays, comprising 57 patients each. Error bars = standard errors of means. 3 3 Mean Degree 2 1 Proper hepatic artery Proximal portal veins Hepatic veins Mean Degree 2 1 Splenic parenchyma Splenic moiré pattern Liver parenchyma A B Fig. 6 Graph showing results of prospective qualitative image review. A, Mean degree of proper hepatic arterial enhancement was constantly high at 5 15 seconds and was higher at 2 seconds than at 25 3 seconds (p <.5). Mean degree of proximal portal venous enhancement increased constantly from 5 25 seconds and plateaued at seconds. It was higher at 1 15 seconds than at 5 seconds (p <.1) and higher at 25 3 seconds than at 2 seconds (p <.1). B, Mean degree of liver parenchyma increased constantly at 5 3 seconds and peaked at 45 seconds. It was higher at 15 seconds than at 5 seconds (p <.5) and higher at 3 seconds than at 2 seconds (p <.5). Mean degree of splenic enhancement moiré pattern was high at 5 15 seconds. Mean moiré degree was significantly higher at 2 seconds than at 3 seconds (p <.5). Note that x-axis is indicative not of time course (repeated measurement) but of three different subgroups with different imaging delays, comprising 57 patients each. Error bars = standard errors of means. AJR:187, July 26 W29

6 Degree of HCC Arterial Enhancement and Washout HCC arterial enhancement HCC washout Liver with chronic liver damage and HCC Fig. 7 Graph showing scan delay after trigger versus qualitative degree of tumor arterial enhancement, washout of hepatocellular carcinoma (HCC), and mean degree of liver with chronic liver damage and HCCs. Mean degree of tumor arterial enhancement was higher at 1 15 seconds than at 5 seconds (p <.1) and higher at 2 seconds than at 25 3 seconds (p <.1). Mean degree of tumor washout was high at seconds, although difference was not significant. Mean degree of liver parenchyma in patients with HCCs constantly increased at images obtained with scan delay of 5 45 seconds and then plateaued. Note that x-axis is indicative not of time course (repeated measurement) but of three different subgroups with different imaging delays, comprising all of 3 patients. Error bars = standard errors of means. A Fig year-old woman with cirrhosis and hypervascular 2-mm-sized hepatocellular carcinoma (HCC) in right hepatic lobe. Amount of contrast material administered was 1 ml. A, Transverse image at level of right lower segment of liver with scan delay of 1 seconds after trigger showing dense contrast enhancement in abdominal aorta (asterisk), moderate enhancement in HCC (arrow), intense enhancement in spleen (arrowhead), and minimal enhancement in liver parenchyma. Note sufficient enhancement of HCC and preferable spleen-to-liver contrast difference, suggesting that this phase is optimal as hepatic artery phase for detecting hypervascular HCCs. B, Transverse image at same level as A with scan delay of 5 seconds after trigger, which shows decreased contrast enhancement in abdominal aorta (asterisk) and spleen (arrowhead), and moderate washout in HCC (large arrow). Note that liver parenchymal enhancement is weak owing to extrahepatic portosystemic shunting (small arrows) caused by portal hypertension. B ied the effect of reduced cardiac output in a porcine model and concluded that the times to arrival of contrast bolus in the aorta and to peak aortic and hepatic enhancement increase as cardiac output decreases. However, a reduction in cardiac output resulted in a substantial increase in peak aortic enhancement but not in peak hepatic enhancement. We found that aortic transit times widely varied from 6 to 33 seconds in our series, which may have reflected the difference in cardiac output of individual patients. Therefore, a bolus-tracking program should be used to compensate for the dispersion of optimal delay times in individual patients. Our quantitative results show that the aorta had a peak enhancement at a scan delay of 1 seconds (CEI of 336 H) after the trigger (density of the aorta reaching 5 H when obtained at the level of the diaphragm) and that the splenic enhancement moiré pattern was prominent at a scan delay of 1 15 seconds after the trigger, although the splenic enhancement at the same time frame (97 18 H) was yet to peak at a scan delay of 2 seconds (117 H). At a scan delay of 1 15 seconds after the trigger, the proximal portal veins were moderately enhanced (75 14 H), but the liver parenchyma and the hepatic veins were negligibly enhanced (15 25 H and H, respectively); however, HCC-toliver contrast and tumor arterial enhancement were maximal at this time. These observations suggest that a scan delay of 1 15 seconds after the trigger is an optimal scan delay for hepatic arterial dominant phase imaging for the detection of hypervascular tumors (Fig. 8A). The hepatic artery phase images obtained at a scan delay of 1 15 seconds after the trigger may correspond to so-called portal venous inflow phase images [27]. The enhancement of liver parenchyma exceeded 5 H at a scan delay of 3 seconds after the trigger and that of the proximal portal veins peaked at a scan delay of 25 seconds af- W3 AJR:187, July 26

7 MDCT of the Liver and Hypervascular Hepatocellular Carcinomas Fig year-old man with cirrhosis and hypervascular-sized hepatocellular carcinoma (HCC) 25 mm in right hepatic lobe. Amount of contrast material administered was 136 ml. Transverse image at level of porta hepatis with scan delay of 5 seconds after trigger shows dense contrast enhancement in abdominal aorta (asterisk) and proximal hepatic arterial branches (small arrows), moderate enhancement in spleen (arrowhead), mild enhancement of HCC (large arrow), and minimal enhancement in liver parenchyma and portal trunk (curved arrow). Note insufficient HCC and spleen enhancement, suggesting that timing is somewhat premature as hepatic arterial phase for detecting hypervascular HCCs, but that this phase is optimal for 3D CT angiography reconstruction. ter the trigger, which suggests that the optimal portal venous phase starts at a scan delay of 3 seconds after the trigger. However, at this time frame aortic and splenic enhancement were still high (179 and 16 H, respectively), which suggests that a 3-second delay after the trigger was premature for the portal venous phase, because hypervascular tumors in the liver might be continuing to enhance at this time and could be obscured. The liver parenchyma was most intensely enhanced (59 63 H) at seconds after the trigger, splenic enhancement was reduced at the same time frame (73 82 H), and the HCC-to-liver contrast was below zero at this time. We infer that the optimal portal venous phase for the detection of hypervascular tumors may start at a scan delay of 45 seconds or even later after the trigger (Fig. 8B). Foley et al. [27] stated that the use of an early arterial phase played an important role in the generation of 3D reconstruction images that facilitated the understanding of the vascular anatomy, although the phase was not optimal for diagnosing hepatic tumors. For 3D displays of the hepatic arteries in preoperative evaluations, images obtained at a scan delay of 5 seconds after the trigger might be optimal because the CEI of the aorta was very intense and the qualitative degree of proper hepatic arterial enhancement was prominent enough, with virtually no enhancement of the proximal portal veins at this time frame (Fig. 9). When simultaneously obtaining early hepatic artery phase images for hypervascular 3D reconstructions and late hepatic artery phase images for tumor detection, radiologists would be able to obtain biphasic images within a single breath-hold by starting the first phase at a scan delay of 5 seconds and the second phase at a scan delay of 1 15 seconds after the trigger, depending on the time interval necessary between the two scans. Observation of hemodynamics in focal hepatic lesions is important in the differential diagnosis of malignant tumors. Murakami et al. [28] recommended double hepatic arterial phase imaging with MDCT for improving the detection of hypervascular HCCs and reducing false-positive lesions. To perform double hepatic arterial phase imaging, one should be able to obtain biphasic images within a single breath-hold by starting the first phase at a scan delay of 1 seconds and the second phase at a scan delay of 15 2 seconds after the trigger, depending on the time interval necessary between the two scans. Even if the portal venous phase imaging is commenced at a scan delay of 45 seconds after the trigger, patients are allowed a breathing interval of at least 2 seconds between the double hepatic arterial and portal venous phase scans. Although we used an injection rate of 4 ml/s in all patients in our study, the contrast injection rate is a topic of debate. Bae et al. [7] reported that injection rates of more than 2 ml/s did not substantially increase peak hepatic enhancement, but they did help increase the magnitude of arterial enhancement and the temporal separation of hepatic arterial and portal venous phases of enhancement on dual-phase helical CT. We believe that an injection rate of 4 ml/s is optimal for the better separation of hepatic arterial and portal venous phases, although the separation might be easier owing to the short acquisition time of MDCT, if a lower injection rate were used. Some limitations of our study should be mentioned. First, our study population included patients with or without chronic liver damage, although HCCs commonly arise in livers with chronic liver damage. In our subanalysis of 3 patients with HCCs, most of whom also had chronic liver damage, the liver parenchyma most intensely enhanced (51 59 H) at seconds, although the enhancement was somewhat weaker with livers in the 3-patient subgroup than with livers overall (59 63 H). Based on these results, we infer that the optimal scan delay for portal venous phase imaging, aiming for HCC detection by revealing tumor washout, may not differ much between livers with and without chronic liver damage. However, severe chronic liver damage causing portal hypertension may well affect the intensity and timing of maximal liver parenchymal enhancement. Second, although a trend to significance was shown in the time density curves (36 for HCC) and the quantitative and qualitative assessments showed agreement, the lack of statistical significance was probably because of the insufficient size of the population of patients with HCC. In the current study, we assessed optimization for the detection of hypervascular HCCs, but not for hypovascular hepatic tumors such as metastases. Portal venous phase imaging, in which the liver is most intensely enhanced and the greatest tumor-to-liver contrast is achieved, should be investigated in a future work. In conclusion, optimal enhancement of the hepatic arterial and portal venous phases is achieved when images are obtained from scans initiated at 1 15 and seconds after the arrival of contrast material in the lower thoracic aorta. This imaging delay should apply when the given iodine concentration (3 mg I/mL, 2 ml/kg of body weight) is injected IV at a rate of 4 ml/s and a threshold of 5 H is used for a bolus-tracking program. AJR:187, July 26 W31

8 References 1. Baron RL, Oliver JH III, Dodd GD III, Nalesnik M, Holbert BL, Carr B. Hepatocellular carcinoma: evaluation with biphasic, contrast-enhanced, helical CT. Radiology 1996; 199: Lee HM, Lu DS, Krasny RM, Busuttil R, Kadell B, Lucas J. Hepatic lesion characterization in cirrhosis: significance of arterial hypervascularity on dual-phase helical CT. AJR 1997; 169: Leslie DF, Johnson CD, MacCarty RL, Ward EM, Ilstrup DM, Harmsen WS. Single-pass CT of hepatic tumors: value of globular enhancement in distinguishing hemangiomas from hypervascular metastases. AJR 1995; 165: Mitsuzaki K, Yamashita Y, Ogata I, Nishiharu T, Urata J, Takahashi M. Multiple-phase helical CT of the liver for detecting small hepatomas in patients with liver cirrhosis: contrast-injection protocol and optimal timing. AJR 1996; 167: Yamashita Y, Komohara Y, Takahashi M, et al. Abdominal helical CT: evaluation of optimal doses of intravenous contrast material a prospective randomized study. Radiology 2; 216: Awai K,Takada K Onishi H, Hori S. Aortic and hepatic enhancement and tumor-to-liver contrast: analysis of the effect of different concentrations of contrast material at multi-detector row helical CT. Radiology 22; 224: Bae KT, Heiken JP, Brink JA. Aortic and hepatic peak enhancement at CT: effect of contrast medium injection rate pharmacokinetic analysis and experimental porcine model. Radiology 1998; 26: Foley WD, Hoffmann RG, Quiroz FA, Kahn CE Jr, Perret RS. Hepatic helical CT: contrast material injection protocol. Radiology 1994; 192: Tublin ME, Tessler FN, Cheng SL, Peters TL, McGovern PC. Effect of injection rate of contrast medium on pancreatic and hepatic helical CT. Radiology 1999; 21: Bae KT. Peak contrast enhancement in CT and MR angiography: when does it occur and why? Pharmacokinetic study in a porcine model. Radiology 23; 227: Stevens WR, Johnson CD, Stephens DH, Batts KP. CT findings in hepatocellular carcinoma: correlation of tumor characteristics with causative factors, tumor size, and histologic tumor grade. Radiology 1994; 191: Lee KH, Choi BI, Han JK, Jang HJ, Kim TK, Han MC. Nodular hepatocellular carcinoma: variation of tumor conspicuity on single-level dynamic scan and optimization of fixed delay times for two-phase helical CT. J Comput Assist Tomogr 2; 24: Bae KT, Heiken JP, Brink JA. Aortic and hepatic contrast medium enhancement at CT. Part I. Prediction with a computer model. Radiology 1998; 27: Bae KT, Heiken JP, Brink JA. Aortic and hepatic contrast medium enhancement at CT. Part II. Effect of reduced cardiac output in a porcine model. Radiology 1998; 27: Shimizu T, Misaki T, Yamamoto K, Sueyoshi K, Narabayashi I. Helical CT of the liver with computer-assisted bolus-tracking technology: scan delay of arterial phase scanning and effect of flow rates. J Comput Assist Tomogr 2; 24: Kim T, Murakami T, Hori M, et al. Small hypervascular hepatocellular carcinoma revealed by double arterial phase CT performed with single breath-hold scanning and automatic bolus tracking. AJR 22; 178: Itoh S, Ikeda M, Achiwa M, et al. Late-arterial and portal-venous phase imaging of the liver with a multislice CT scanner in patients without circulatory disturbances: automatic bolus tracking or empirical scan delay? Eur Radiol 24; 14: Declaration of Helsinki. Recommendations guiding doctors in clinical research. Adopted by the World Medical Association in Wis Med J 1967; 66: Fleiss JL. The analysis of variance and multiple comparisons. In: Fleiss JL, ed. The design and analysis of clinical experiments. New York, NY: Wiley, 1986: Hollete MD, Jeffrey RB Jr, Nino-Murcia M, Jorgensen MJ, Harris DP. Dual-phase helical CT of the liver: value of arterial phase scans in the detection of small ( 1.5 cm) malignant hepatic neoplasms. AJR 1995; 164: Bonaldi VM, Bret PM, Reinhold C, Atri M. Helical CT of the liver: value of an early hepatic arterial phase. Radiology 1995; 197: Oliver JH III, Baron RL, Federle MP, Rockette HE Jr. Detecting hepatocellular carcinoma: value of unenhanced or arterial phase CT imaging or both used in conjunction with conventional portal venous phase contrast enhanced CT imaging. AJR 1996; 167: Hwang GJ, Kin MJ, Yoo HS, Lee JT. Nodular hepatocellular carcinomas: detection with arterial-, portal-, and delayed-phase images at spiral CT. Radiology 1997; 22: Oliver JH III, Baron RL. Helical biphasic contrast-enhanced CT of the liver: technique, indications, interpretations, and pitfalls. Radiology 1996; 21: Takayasu K, Furukawa H, Wakao F, et al. CT diagnosis of early hepatocellular carcinoma: sensitivity, findings, and CT pathologic correlation. AJR 1995; 164: Kanematsu M, Oliver JH, Carr B, Baron RL. Hepatocellular carcinoma: the role of helical biphasic contrast-enhanced CT versus CT during arterial portography. Radiology 1997; 25: Foley WD, Mallisee TA, Hohenwalter MD, Wilson CR, Quiroz FA, Taylor AJ. Multiphase hepatic CT with a multirow detector CT scanner. AJR 2; 175: Murakami T, Kim T, Takamura M, et al. Hypervascular hepatocellular carcinoma: detection with double arterial phase multi-detector row helical CT. Radiology 21; 218: W32 AJR:187, July 26