Detection of Small Pancreatic Tumors with Multiphasic Helical CT

Transcription

1 Downloaded from by on 11/24/17 from IP address opyright RRS. For personal use only; all rights reserved Yulia Lisenko ronstein 1 Evelyne M. Loyer 2 Harmeet Kaur 2 Haesun hoi 2 ynthia David 2 Ronelle. Durow 2 Lyle D. roemeling 3 Karen R. leary 4 husilp harnsangavej 2 Received March 26, 2003; accepted after revision September 16, Department of Radiology, Lady Davis Medical enter, Michal St. 7, Haifa 34362, Israel. 2 Department of Diagnostic Radiology, The University of Texas M. D. nderson ancer enter, 1515 Holcombe lvd., Houston, TX ddress correspondence to E. M. Loyer. 3 Department of iostatistics, The University of Texas M. D. nderson ancer enter, Houston, TX Department of Pathology, The University of Texas M. D. nderson ancer enter, Houston, TX JR 2004;182: X/04/ merican Roentgen Ray Society Detection of Small Pancreatic Tumors with Multiphasic Helical T OJETIVE. The purpose of this study was to evaluate the sensitivity and specificity of helical T in the detection of adenocarcinomas of the pancreas measuring 2 cm or smaller at pathologic examination. MTERILS ND METHODS. Thin-section triple phase (20, 40, and 70 sec after the start of injection) contrast-enhanced helical T scans of the abdomen in 18 patients with a pancreatic carcinoma that was 2 cm or smaller and 18 patients with a normal pancreas were retrospectively reviewed by two senior radiologists who specialized in oncologic abdominal imaging. Discrepancies were resolved by consensus. The observers were unaware of the clinical information. T scans were evaluated for the presence of a pancreatic mass, bile, and pancreatic duct stricture. The location and size of tumors as determined on T were compared with pathologic findings. The T results were also compared with the prospective T interpretations derived from the radiology reports and with the endoscopic sonographic reports when available. RESULTS. The sensitivity of thin-section triple-phase helical T in the detection of small pancreatic masses was 77%, and the specificity was 100% for the two experienced observers. The sensitivity and specificity were 72% and 100%, respectively, for the prospective interpretations done by 10 observers. There was no correlation between the tumor size at pathology and the T measurements. ONLUSION. Thin-section contrast-enhanced helical T is sensitive and highly specific for the detection of pancreatic tumors measuring 2 cm or smaller. Improvement in the detection rate of this technique compared with previous techniques lies in the optimization of parenchymal enhancement during the pancreatic phase and the decrease in slice thickness. T he 5-year survival rate for patients with pancreatic duct adenocarcinoma is less than 5% for all stages combined [1]. In patients with resectable carcinoma of the pancreatic head, a long-term survival rate of 20% can be achieved with a combination of surgery, chemotherapy, and radiation therapy [2]. Resectability is determined from imaging and is based on the absence of extrapancreatic involvement, presence of a patent superior mesenteric vein portal vein confluence, and absence of direct extension of tumor to the celiac axis or superior mesenteric artery [2]. lthough T and MRI perform equally well in the detection of pancreatic carcinoma [3], the availability of helical T and technologic advances have made it the method of choice for pancreatic imaging [3, 4]. urrent reported data on the sensitivity of T in the detection of pancreatic adenocarcinoma stem from studies in which the mean tumor size was greater than 2 cm and in which T was performed with technical parameters that would now be considered suboptimal [3, 5 9]. survival benefit from surgery is achieved only in patients who undergo a negative-margin pancreaticoduodenectomy [2]. The fact that increasing tumor size correlates with an increasing rate of unresectability and decreasing survival rate [10] underscores the need to detect tumors while they are small and have not spread locally. This need prompted us to evaluate the sensitivity and specificity of a triplephase contrast-enhanced helical T protocol for the detection of small ( 2 cm in greatest dimension) pancreatic adenocarcinomas. Materials and Methods Patient Selection etween January 1997 and June 2000, 94 consecutive patients underwent pancreaticoduodenec- JR:182, March

2 ronstein et al. Downloaded from by on 11/24/17 from IP address opyright RRS. For personal use only; all rights reserved tomy for pancreatic carcinoma at our institution. The patients were identified from our pancreatic tumor database. The initial diagnoses were based on T or on findings at endoscopic sonographically guided biopsy. ll the patients received preoperative chemoradiation that included a combination of external beam radiation therapy and 5-fluorouracil. The patients underwent restaging after chemoradiation and then pancreaticoduodenectomy within 4 53 days of restaging (mean, 14.1 days). Fourteen patients with tumors measuring cm (mean, 1.66 cm) at pathologic examination and four patients with only microscopic foci of tumor were included in our study. There were 12 men and six women who ranged in age from 39 to 78 years (mean, 62.6 years). The tumor was located in the pancreatic head or uncinate process in all 18 patients. In 14 patients, a biliary stent had been placed into the common bile duct before imaging. control group composed of 18 patients with no suspected pancreatic disease who had undergone the same triple-phase T protocol was identified from the T logbooks. There were seven men and 11 women who ranged in age from 40 to 78 years (mean, 58.1 years). mong these were patients with hepatic metastasis or primary liver tumors and patients with extrahepatic, extrapancreatic tumors. Patients who had undergone hepatic resection or had massive hepatic involvement by cancer were excluded. In two patients, a biliary stent was present. T Protocol ll T scans were obtained using either a LightSpeed multidetector scanner or a HiSpeed dvantage single-detector scanner (both General Electric Medical Systems, Milwaukee, WI). arium sulfate (2%) (Readi-at 2, E-Z-EM, Westbury, NY) was administered orally. Unenhanced helical scans through the liver and pancreas were acquired with a pitch of 1 and slice thickness of 10 mm on the single-detector units. Unenhanced helical scans through the liver and pancreas were acquired with a pitch of 6 and slice thickness of 10 mm on the multidetector units. ontrast-enhanced helical scans of the pancreas were acquired in a craniocaudal direction, during a single breathhold, using 3-mm slice thickness and a pitch of 2 on the single-detector units. ecause of limited anatomic coverage with a single-detector unit, the examination was designed to encompass the entire pancreas and to cover as much liver parenchyma as possible in a single breath-hold while at the same time constantly maintaining the parameters set for the pancreas imaging (slice, delay). onsequently, with the single-detector unit, the entire liver was not imaged during the arterial and parenchymal phases. With the multidetector units, contrast-enhanced scans of the entire liver and pancreas were acquired in a craniocaudal direction with a single breath-hold using 5-mm slice thickness and a pitch of 6. The multidetector slices were retrospectively reconstructed at 2.5-mm thickness. There were 11 patients evaluated with the single-detector unit and seven patients evaluated with the multidetector unit. The scanning sequences were initiated 20 sec (arterial phase), 40 sec (pancreatic phase), and 70 sec (portal venous phase) after the start of IV injection of 150 ml of 60% nonionic contrast material (Optiray 320 [ioversol], Mallinkrodt, St. Louis, MO) at a rate of 5 ml/sec. The scanning time for each phase was sec. Scans were obtained at 120 kv and m. Scans were printed with both narrow and soft-tissue window settings. Interpretation of Images Each T study was reviewed in a retrospective blinded fashion by two radiologists who specialize in oncologic abdominal imaging. Discrepancies over the presence of a measurable tumor were encountered in two cases and were resolved by consensus. Only the study performed at the completion of chemoradiation, just before surgery, was analyzed. Each scan was assessed for the presence of a pancreatic tumor on the basis of the detection of a focal abnormality in attenuation or texture or a contour deformity, associated or not with an abrupt change in the caliber of the bile or pancreatic duct. hange in texture is defined as a loss of the lobulation usually seen in the normal pancreatic parenchyma. The location and size of the presumed tumor were recorded. The observers were not asked to assess resectability or to evaluate for liver, nodal, or peritoneal metastasis. The false-negative studies were reviewed a second time by a different radiologist with knowledge of the surgical and pathologic findings. The original prospective interpretations of the studies derived from the radiology reports were reviewed and compared with the retrospective evaluations for the detection of tumors. total of 10 radiologists had participated in the prospective interpretation. Endoscopic Sonography Endoscopic sonography was performed in 10 patients within 10 days of the initial diagnostic T. Endoscopic sonography reports were reviewed and correlated with the T interpretations and the pathology reports. Pathologic Data Pathology reports were reviewed in respect to size, location, appearance, and histologic features of the mass and of the normal pancreatic parenchyma. For the T studies with false-negative findings, the slides were reviewed a second time. size of 0 cm was recorded when numerous microscopic tumor foci were reported scattered in the pancreas without a discrete mass. The status of the surgical margin, which was invariably negative for malignancy, was also noted. Statistical nalysis The sensitivity and specificity of this protocol for the detection of small pancreatic tumors were calculated. Positive predictive values were calculated separately for each radiographic criterion indicative of tumor: attenuation, texture, or contour abnormality and ductal obstruction. Observer s preference relative to the phase of contrast enhancement was analyzed. Pearson s correlation between the tumor size reported by the pathologist and that radiologically assessed by the observers was also performed. Results tumor was identified on T scans by the two observers in 14 of the 18 patients with pancreatic cancer (Figs. 1 and 2). Discrepancy over the presence of a mass occurred in two cases. In both instances, disagreement was resolved by consensus, with both observers recognizing the presence of a mass. The sensitivity for the detection of small tumors was therefore 78% (14/18). The remaining four T studies were interpreted as negative for the presence of a measurable tumor. In the control group, all 18 studies were interpreted as negative for the presence of a pancreatic tumor by the two observers, giving a specificity of 100%. The studies with false-negative findings included two patients in whom pathologic examination showed only minimal microscopic residual tumor but no discrete mass. In one of these patients, endoscopic sonography was available and revealed a 1.3-cm mass. In the remaining two false-negative cases, a mass measuring 2 cm was found in both instances at pathologic examination. In both cases, the tumor involved the head of the pancreas with extension into the peripancreatic adipose tissue. On T, both patients had strictures of the distal bile duct with stents in place. These studies were interpreted as bile duct strictures without definite mass or evidence of pancreatitis, to be further evaluated on ERP and endoscopic sonography. In both cases, results of endoscopic sonography were negative (Fig. 3). There was no correlation between the tumor size measured by the radiologists and the tumor size reported by the pathologists, nor was there correlation between endoscopic sonography and pathology measurements or between endoscopic sonography and T measurements. The tumor size varied from 1 to 3 cm by T measurement and from microscopic foci to 2 cm at pathology. The lack of correlation between the different measurements did not follow a pattern. Review of the original radiology reports showed that prospectively 10 studies were in- 620 JR:182, March 2004

3 Downloaded from by on 11/24/17 from IP address opyright RRS. For personal use only; all rights reserved T of Small Pancreatic Tumors terpreted as positive for tumor, three as suspicious for a mass, and five as negative for a mass. Only one of the false-negative studies was interpreted as positive for tumor during the retrospective interpretation. The remaining four studies were also interpreted as negative in the retrospective study. The sensitivity of the technique in this setting was 72% (13/18) and specificity was 100%. The radiographic signs most helpful for the diagnosis of a pancreatic tumor were a focal area of low attenuation or a change in texture at the site of an abrupt interruption of the bile or pancreatic duct (Figs. 1 and 2). There were individual variations in the radiographic analysis: a texture abnormality was detected in 13 of 14 cases by one observer and in only two cases by the other observer. Other findings such as focal hyperattenuation (1/14 cases) and contour abnormality (2/14 cases) were rarely observed (Figs. 1 and 4). Fig year-old man with small hypodense mass that exhibits typical appearance of pancreatic carcinoma., ontrast-enhanced T scan shows small hypodense mass (arrow) deforming contours of inferior portion of head of pancreas and dilated common bile duct (arrowhead)., ontrast-enhanced T scan obtained below level of shows mass (arrow) and abrupt obstruction of common bile duct (arrowhead)., ontrast-enhanced T scan obtained below level of shows small hypodense mass (arrow). Fig year-old woman with small hypodense mass that exhibits typical appearance of pancreatic carcinoma., ontrast-enhanced T scan shows obstructed pancreatic and common bile duct (arrowhead) with atrophy of pancreatic body (arrow). iliary stent is in place., ontrast-enhanced T scan obtained below level of shows hypodense mass (arrow) in head of pancreas. Pancreatic parenchyma has lost its normal lobular texture at site of mass. iliary stent is seen within mass. Fig year-old man with false-negative T findings for pancreatic carcinoma. Tumor was in head of pancreas and measured 2 cm at pathologic examination., ontrast-enhanced T scan shows dilated common bile duct (arrowhead) with stent in place. Pancreatic duct and pancreatic parenchyma are normal., ontrast-enhanced T scan obtained below level of shows normal pancreatic head. iliary stent (arrowhead) is in place in common bile duct., ontrast-enhanced T scan obtained below level of shows normal head and ampullary region. iliary stent is seen in ampullary region. JR:182, March

4 Downloaded from by on 11/24/17 from IP address opyright RRS. For personal use only; all rights reserved ronstein et al. Fig year-old man with pancreatic carcinoma. Isodense mass in head of pancreas can be detected by change in texture., ontrast-enhanced T scan shows obstructed pancreatic duct (arrowhead)., ontrast-enhanced T scan obtained below level of shows isodense mass (arrow) in head of pancreas. Pancreatic parenchyma has lost its normal lobular texture at site of mass., ontrast-enhanced T scan obtained below level of shows normal-appearing parenchyma in which lobulation can be defined (arrow). Resolving inflammation is seen along superior mesenteric artery (arrowheads). The pancreatic phase was considered the optimal phase for tumor detection in 12 of 14 patients by one observer and in nine of 14 patients by the other observer. Discussion Surgical resection is currently the only potentially curative treatment for pancreatic carcinoma. This therapeutic approach is possible when the tumor does not involve vascular structures that cannot be resected and when it has not metastasized [2]. study conducted by riyama et al. [10] showed a 100% 5-year postoperative survival rate for patients whose tumors were smaller than 1 cm and limited to the intraductal epithelium without parenchymal, vascular, perineural, or lymphatic invasion. However, the survival rate decreased significantly as tumor size increased beyond 1 cm. The detection of small pancreatic tumors can be challenging. Of the imaging techniques that are available, T is the most commonly used. Detection is based primarily on the difference in enhancement between the tumor and normal parenchyma. Most ductal adenocarcinomas have an abundant, dense fibroblastic stroma with a decreased number of vessels within the tumor [2]. onsequently, these tumors enhance less than the surrounding normal parenchyma. dditional radiographic findings may include a focal change in texture; abrupt change in the caliber of the pancreatic duct, bile duct, or both, particu- Fig year-old man with pancreatic carcinoma. Pathologically detected microscopic foci of tumor were interpreted as mass on contrast-enhanced T., ontrast-enhanced T scan shows stent in common bile duct. ephalad portion of head (arrow) is small., ontrast-enhanced T scan below level of shows small, low-attenuation focus (arrow) seen medial to stent was interpreted as tumor. Low-attenuation focus is associated with enlargement of caudal aspect of head. 622 larly when associated with a density or texture change; and modification of the contours of the pancreas (Figs. 1 and 2). The sensitivity of dual-phase contrast-enhanced helical T in the detection of pancreatic carcinoma has been reported to vary from 76% to 92% [3, 5, 6, 9]. However, the reported detection of small tumors has been poor, with a sensitivity of 67% for tumors smaller than 1.5 cm in one study [6] and 58% for tumors smaller than 2 cm in another study [5]. In our study, triple-phase contrast-enhanced T performed on a state-of-the-art MDT or a single-detector helical T scanner with a contrast injection rate of 5 ml/sec had a sensitivity of 77% and specificity of 100% for the detection of pancreatic cancers smaller than 2 cm in diameter. There were four false-negative studies: two cases in which pathologic examination did not identify a discrete mass but found only microscopic foci of tumor cells, and two in which masses each measuring 2 cm were identified pathologically. In the latter two cases, endoscopic sonography evaluations and T scans were interpreted as negative for a mass. In both cases, a stricture of the distal common bile duct appeared isolated (Fig. 3). Review of the T studies with knowledge of the pathologic findings did not change the false-negative interpretation for a mass and did not explain the discrepancy between the imaging and pathologic findings. Interestingly, there were two other cases with only microscopic disease pathologically, and these were interpreted as positive for tumor on T in both and positive for tumor on endoscopic sonography in one of them. These T diagnoses were based on a subtle change in texture JR:182, March 2004

5 T of Small Pancreatic Tumors Downloaded from by on 11/24/17 from IP address opyright RRS. For personal use only; all rights reserved or density or an abrupt change in the size of the pancreatic duct. One could argue that the setting of the study lent itself to overinterpretation (Fig. 5). lthough specificity was high in our study, it may actually be lower, because there was no patient in our study group with the diagnosis of focal chronic pancreatitis, a condition that may mimic carcinoma on imaging [11]. No correlation existed between the size of the tumor measured pathologically and the size of the tumor measured on T. One likely reason is the difficulty in determining tumor size pathologically. fter chemoradiation, the pancreas is usually fibrotic, making gross detection of the tumor difficult. Lack of correlation was also evident between endoscopic sonography and pathology and between endoscopic sonography and T. Endoscopic sonography, however, was performed at the time of diagnosis, before chemoradiation, making the size comparison less rigorous. We do not have a clear explanation for these disparities and can only assume that the difficulties arise from the poor definition and heterogeneity of the tissue studied. The number of cases in our studies is not sufficient to compare the sensitivity of endoscopic sonography with that of T. s expected, the radiographic findings most helpful for determining the presence of a tumor were a focal area of hypoattenuation, a change in texture associated with an abrupt change in the bile or pancreatic duct caliber, or both. change in contour was rarely observed because of the small size of the tumors. The presence of a biliary stent is a source of problems because the exact site of obstruction may be obscured if the obstruction is affecting the bile duct only and because of intercurrent inflammation around the stent that can be confused with tumor. Optimally, imaging should be performed before stent placement. Falsenegative studies were characterized by isolated stricture of the bile duct, pancreatic duct, or both without texture or density change. In summary, the tumor diagnosis was based on the observation of one or more of these radiographic findings, but there was notable interobserver variation in the perception and analysis of the findings, particularly in the visualization of a texture change. We think that this interobserver variation is because change in texture is perceived more subjectively and was given a different weight by each of the observers. The pancreatic phase of enhancement was rated superior to the arterial or portal venous phases. The poor performance of the arterial phase has previously been reported [6 9, 12]. lthough in our study the pancreatic phase was preferred to the portal phase, McNulty et al. [4], using MDT, found the pancreatic and portal phases to be equivalent. This difference is not explained by technical differences because our protocol is very similar to theirs. possible explanation is that the thinner slice thickness in the pancreatic phase positively affected the observers confidence in our study, particularly because the mean tumor size was 1.66 cm in our study versus 3.3 cm in the study by McNulty et al. In our study, observers were not asked to specifically compare the phases of enhancement but only to rate the one they judged the best. The quantitative assessment by McNulty et al. indicates that although pancreatic enhancement is maximal during the pancreatic phase, tumor conspicuity is equivalent in the pancreatic and portal phases. higher injection rate, optimal parenchymal opacification, and thinner slice thickness achieved by a decrease in scanning time are the main technical differences between our protocol and those described in the literature [3, 5 9], with the exception of the study by McNulty et al. [4]. T protocols to detect and stage pancreatic tumors are designed to optimize enhancement of the pancreas to ensure maximal tumor-versus-pancreas contrast and to optimize enhancement of the peripancreatic vessels and liver for staging. Studies on pancreatic enhancement have defined general principles to follow in imaging the pancreas. Peak pancreatic enhancement, which defines the pancreatic phase, occurs before peak enhancement of the liver [12 14]. The time when peak pancreatic enhancement occurs varies with the rate of injection [11, 13]. Enhancement is also affected by the amount of iodine used [11]. With single-detector T, the arterial phase is used to assess the vascular anatomy for local staging and the portal phase to detect hepatic metastasis. Experience with MDT, however, is bringing slight modifications to the protocols used to image the pancreas. The shorter scanning time with MDT provides a pancreatic phase and a portal phase that are focused on maximal arterial enhancement and maximal venous enhancement, respectively. In this setting, a pure arterial phase becomes obsolete [4]. In this study, we have shown that triple phase contrast-enhanced helical T is a sensitive method for the detection of small, potentially resectable pancreatic tumors. Our study was done with a triple-phase protocol similar to that reported by McNulty et al. [4]. We and they now use a biphasic technique that provides the same tumor-versus-pancreas contrast as the triple-phase technique but allows thinner collimation and multiplanar reconstruction when needed. References 1. merican ancer Society. ancer facts & figures tlanta, G: merican ancer Society, Evans D, bbruzzese JL, Willet G. ancer of the pancreas. In: Devita VT, Hellman S, Rosenberg S, eds. ancer: principles & practice of oncology. Philadelphia, P: Lippincott Williams & Wilkins, 2001: Sheridan M, Ward J, Guthrie J, et al. Dynamic contrast-enhanced MR imaging and dual-phase helical T in the preoperative assessment of suspected pancreatic cancer: a comparative study with receiver operating characteristic analysis. JR 1999;173: McNulty NJ, Francis IR, Platt JF, ohan RH, Korobkin M, Gebremariam. Multi-detector row helical T of the pancreas: effect of contrast-enhanced multiphasic imaging on enhancement of the pancreas, peripancreatic vasculature, and pancreatic adenocarcinoma. Radiology 2001;220: Ichikawa T, Haradome H, Hachiya J, et al. Pancreatic ductal adenocarcinoma: preoperative assessment with helical T versus dynamic MR imaging. Radiology 1997;202: Legmann P, Vignaux O, Dousset, et al. Pancreatic tumors: comparison of dual-phase helical T and endoscopic sonography. JR 1998;170: Lu DS, Vedantham S, Krasny RM, Kadell, erger WL, Reber H. Two-phase helical T for pancreatic tumors: pancreatic versus hepatic phase enhancement of tumor, pancreas, and vascular structures. Radiology 1996;199: oland GW, O Malley ME, Saez M, Fernandez-del- astillo, Warshaw L, Mueller PR. Pancreatic-phase versus portal vein-phase helical T of the pancreas: optimal temporal window for evaluation of pancreatic adenocarcinoma. JR 1999;172: Tabuchi T, Itoh K, Ohshio G, et al. Tumor staging of pancreatic adenocarcinoma using early- and late-phase helical T. JR 1999;173: riyama J, Suyama M, Satoh K, Sai J. Imaging of small pancreatic ductal adenocarcinoma. Pancreas 1998;16: Kim T, Murakami T, Takamura M, et al. Pancreatic mass due to chronic pancreatitis: correlation of T and MR imaging features with pathologic findings. JR 2001;177: Hollett MD, Jorgensen MJ, Jeffrey R Jr. Quantitative evaluation of pancreatic enhancement during dual-phase helical T. Radiology 1995;195: Tublin ME, Tessler FN, heng SL, Peters TL, McGovern P. Effect of injection rate of contrast medium on pancreatic and hepatic helical T. Radiology 1999;210: Kim T, Murakami T, Takahashi S, et al. Pancreatic T imaging: effects of different injection rates and doses of contrast material. Radiology 1999;212: JR:182, March