Alex M. Aisen

Comparison of Endoscopic Ultrasonography and Multidetector Computed Tomography for Detecting and Staging Pancreatic Cancer

Context Clinicians often use multidetector computed tomography or endoscopic ultrasonography to detect and stage pancreatic cancer. Contribution This prospective study found that, among 80 adults with proven pancreatic cancer, the sensitivity of multidetector computed tomography and endoscopic ultrasonography for detecting a pancreatic mass was 86% (CI, 77% to 93%) and 98% (CI, 91% to 100%), respectively. Among 53 patients undergoing surgery, endoscopic ultrasonography was more accurate for staging local tumor spread, but both tests showed similar accuracy for nodal staging and detecting resectability. Cautions Optimal strategies to detect and stage pancreatic cancer may vary across sites depending on the expertise of radiologists and endosonographers. The Editors In the United States for the year 2003, it was estimated that pancreatic cancer would be diagnosed in approximately 30700 patients and contribute to 30000 deaths (1). Complete surgical removal with negative histologic margins (R0 resection) is an independent predictor of postoperative survival (2-4) and remains the only potential curative treatment for pancreatic cancer. At surgical exploration, however, only 5% to 25% of the tumors are amenable to resection (5-8). Therefore, the principle goal of preoperative evaluation is to identify patients with potentially resectable disease while avoiding surgical exploration in those with unresectable disease. There is no evidence-based consensus on the optimal preoperative imaging assessment of patients with suspected pancreatic cancer. Because of widespread availability, helical computed tomography (CT) is usually the initial study for this indication (9, 10). Dual-phase helical CT, during which postinjection contrast image acquisition is obtained in both the pancreatic (arterial) and portal venous phases, has improved detection rate and assessment of resectability in patients with suspected pancreatic cancer (11, 12). Current state-of-the-art CT imaging uses a multiple-row detector with narrow detector collimation, wide x-ray beam, and rapid table translation; these features offer faster acquisition and thinner image slices compared with single-detector CT (13-15). Whether multidetector CT offers improved detection and staging of pancreatic cancer, however, is unknown. Endoscopic ultrasonography has been shown to be superior to conventional CT for the detection (16-20) and staging (19, 20) of pancreatic cancer. When compared with helical CT, however, endoscopic ultrasonography is reported to be either equivalent for detection (21, 22) or superior for detection or staging (23-25). To date, no comparative studies of multidetector CT with other imaging tests, including endoscopic ultrasonography, for suspected pancreatic cancer have been performed. Therefore, we conducted a prospective trial to compare endoscopic ultrasonography and multidetector CT for the detection, staging, and resectability of suspected locoregional pancreatic cancer. Methods Patients The institutional review board at Indiana University Medical Center approved this study, and all patients signed written informed consent. Eligible patients were referred to our hospital with clinically suspected or recently diagnosed solid or cystic pancreatic cancer within the previous 8 weeks. The referral base for our hospital consists of gastroenterologists and surgeons from Indiana and the surrounding contiguous states. Patients were eligible only if they agreed to undergo endoscopic ultrasonography, CT, and surgery (if necessary) at our institution. Patients were excluded if they had previously undergone endoscopic retrograde cholangiopancreatography or endoscopic ultrasonography at our institution for suspected pancreatic cancer; declined or remained undecided about potential surgical intervention; were referred to our institution by surgeons outside our hospital system. Patients were also excluded if they were pregnant, were incarcerated, could not independently provide informed consent, or were considered high surgical risk (American Society of Anesthesiology class III to V). In addition, we excluded patients with known or suspected periampullary masses, cholangiocarcinomas, or cancer with suspected locally advanced arterial (superior mesenteric, hepatic, or celiac) involvement or metastatic disease (ascites, suspicious liver or pulmonary lesions, distant enlarged lymph nodes) detected by previous imaging studies. Patients with suspected nonocclusive involvement of the superior mesenteric vein or portal vein were considered eligible for enrollment. Study Design This was a prospective, single-center, observational study. All enrolled patients had to respond to an initial health and medical questionnaire, which was followed by same-day endoscopic ultrasonography. Computed tomography was performed within 1 week. Within 3 weeks after CT, a surgeon examined the patient and reviewed the results of endoscopic ultrasonography and CT to determine eligibility for potential resection. After surgery or the decision to pursue nonoperative management, we telephoned patients to assess quality of life at 1 month, 3 months, and every 6 months until death or until 24 months if clinical disease remained stable. Endoscopic Ultrasonography Technique Conscious sedation was performed with various combinations of intravenously administered propofol, meperidine, fentanyl, or midazolam. Initially, we examined all patients with a radial echoendoscope (Olympus GF-UM130 [Olympus America Inc., Melville, New York]). We then examined patients with a linear echoendoscope (using either Pentax FG-36UX [Pentax Precision Instruments, Orangeburg, New York] or Olympus GF-UC140P [Olympus America Inc.]). Unless cancer had been definitively confirmed previously, endoscopic ultrasonographyguided fine-needle aspiration was performed with a 22-gauge needle (Wilson-Cook Medical Inc., Winston-Salem, North Carolina) in all patients, when applicable. A cytotechnologist or cytopathologist was on-site for preliminary interpretations of all aspirations. One of 3 experienced gastroenterologists, each of whom had performed at least 1000 pancreatic examinations, performed all procedures. The operator was not blinded to previous radiographic data. Recorded information included the presence or absence, size, echocharacteristics, location, or locoregional extension of any visualized pancreatic mass, lymph nodes, or distant metastases. Lymph nodes that were not accessible to endoscopic ultrasonographyguided fine-needle aspiration were considered malignant if 3 or more of the following criteria were present: diffuse hypoechoic echogenicity, short-axis diameter of 5 mm or greater, well-defined borders, round shape, or location within 5 mm of the tumor. Well-defined hypoechoic or hyperechoic lesions within the liver with a short-axis diameter of 10 mm or greater and not accessible to fine-needle aspiration were defined as metastases. We considered vascular involvement by the tumor to be present if any 1 of the following were noted: loss of the normal hyperechoic interface between tumor and vessel for at least 5 mm (adherence), irregular tumor and vessel interface, tumor within vessel lumen (invasion), vessel encasement, and perigastric or periduodenal collaterals with associated venous occlusion. Immediately after the examination, any visualized mass was designated by the endosonographer as surgically resectable or unresectable and assigned a tumor, node, metastasis (TNM) staging according to the 1997 American Joint Committee on Cancer (AJCC) classification (Appendix Table) for staging of pancreatic cancer (26). Multidetector CT Technique We performed multidetector CT with a quad-channel scanner (MX 8000 Quad, Philips Medical Systems, Cleveland, Ohio) by using 0.5-second gantry rotation time and acquisition of 4 sections per rotation. All patients drank 500 mL of tap water as nonopaque oral intraluminal contrast media. A total of 150 mL (300 mg of iodine/mL) of low-osmolar contrast media (Isovue-300, Bracco Diagnostics, Princeton, New Jersey) was injected with a power injector (CT Envision Injector, Medrad, Pittsburgh, Pennsylvania) at a rate of 4.0 mL/s into an antecubital vein by using either an 18- or 20-gauge cannula. Examination was performed in a dual-phase mode. Image acquisition was first done during the pancreatic phase (35 seconds after the start of contrast infusion) from the top to the bottom of the pancreas with 4.0-mm beam collimation (nominal section thickness, 1.0 mm; effective section thickness, 1.3 mm), 0.5-mm reconstruction interval, 120 kVp, 205 mAs, and a pitch of 1.0 during a single breath-hold of 15 to 20 seconds. The second phase was performed during the portal venous phase (65 seconds after the start of contrast infusion) from the top of the liver to the iliac crests with 10-mm beam collimation (nominal section thickness, 2.5 mm; effective section thickness, 3.2 mm), 1.3-mm reconstruction interval, 120 kVp, 250 mAs, and a pitch of 0.875 during a single breath-hold of 15 seconds. Multiplanar (2-dimensional) reformatting was not routinely performed; however, when it was used, the entire data set was transferred to a workstation (MX View, Philips Medical Systems) for evaluation. No 3-dimensional (volume rendering) postprocessing was used in this study. One of 3 experienced gastrointestinal radiologists who were blinded to the results of the previous endoscopic ultrasonography examination interpreted all scans. Patient information provided to the interpreting radiologist included presenting symptoms; the size, location, and vascular involvement (if known) of any visualized pancreatic mass from previous CT; and the results of any previous endoscopic retrograde cholangiopancreatography (for example, presence or absence of ductal strictures) or pathology (for example, endoscopic brush cytology). Locoregional and distant adenopathy were considered malignant if they were greater than 10 mm in

ASSERT: a physician-in-the-loop content-based retrieval system for HRCT image databases

It is now recognized in many domains that content-based image retrieval from a database of images cannot be carried out by using completely automated approaches. One such domain is medical radiology for which the clinically useful information in an image typically consists of gray level variations in highly localized regions of the image. Currently, it is not possible to extract these regions by automatic image segmentation techniques. To address this problem, we have implemented a human-in-the-loop (a physician-in-the-loop, more specifically) approach in which the human delineates the pathology bearing regions (PBR) and a set of anatomical landmarks in the image when the image is entered into the database. To the regions thus marked, our approach applies low-level computer vision and image processing algorithms to extract attributes related to the variations in gray scale, texture, shape, etc. In addition, the system records attributes that capture relational information such as the position of a PBR with respect to certain anatomical landmarks. An overall multidimensional index is assigned to each image based on these attribute values.

Usefulness of diffusion-weighted imaging in the evaluation of renal masses.

OBJECTIVE The objective of our study was to assess the value of diffusion-weighted imaging in differentiating among the various subgroups of renal masses. MATERIALS AND METHODS This retrospective study measured the apparent diffusion coefficients (ADCs) of renal masses. Malignant lesions were confirmed with surgical pathology results. Benign cystic lesions were stable without treatment for a minimum follow-up of 24 months. RESULTS There were 20 and 22 patients, respectively, with benign lesions (three abscess, 31 cysts) and malignant lesions (17 clear cell, five papillary, one chromophobe, and two transitional cell cancers). The malignant lesions were larger than the benign lesions (mean diameter, 4.2 vs 2.6 cm, respectively; p = 0.01, Students t test). The ADC values of the benign lesions were significantly higher than those of the malignant lesions (mean, 2.72 vs 1.88 x 10(-3) mm(2)/s; p < 0.0001). The ADCs of the 31 benign cysts were significantly higher than those of the seven cystic renal cancers (2.77 vs 2.02 x 10(-3) mm(2)/s; p < 0.001). There was no significant difference between the ADCs of clear cell cancers and non-clear cell cancers (1.85 vs 1.97 x 10(-3) mm(2)/s; p = 0.18), but an ADC of less than 1.79 x 10(-3) mm(2)/s was seen only with clear cell cancer. The ADCs of high-grade clear cell cancers (Fuhrman grades III and IV) tended to be lower than those of low-grade clear cell cancers (1.77 vs 1.95 x 10(-3) mm(2)/s; p = 0.12). Among the clear cell cancers, an ADC value of greater than 2.12 x 10(-3) mm(2)/s was seen only with low-grade histology. For differentiating benign from malignant lesions, receiver operating characteristic (ROC) analysis showed an area under the ROC curve of 0.989 (95% CI, 0.919-0.996; p < 0.0001). CONCLUSION ADC measurements may aid in differentiating among the various subgroups of renal masses, particularly benign cystic lesions from cystic renal cell cancers.

Value of Diffusion-Weighted MRI for Assessing Liver Fibrosis and Cirrhosis

OBJECTIVE The objective of our study was to determine the usefulness of the apparent diffusion coefficient (ADC) of liver parenchyma for determining the severity of liver fibrosis. MATERIALS AND METHODS This study investigated 78 patients who underwent diffusion-weighted imaging (DWI) with 1.5-T MRI and pathologic staging of liver fibrosis based on biopsy. DWI was performed with b values of 50 and 400 s/mm(2). ADCs of liver were measured using 2.0- to 3.0-cm(2) regions of interest in the right and left lobes of the liver; the mean ADC value was used for analysis. Pathologic METAVIR scores for liver fibrosis stage were used as a reference standard. RESULTS The mean ADC values for fibrosis pathologically staged using the METAVIR classification system as F0 (n = 11), F1 (n = 16), F2 (n = 10), F3 (n = 14), and F4 (n = 27) were 125.9, 105.0, 104.5, 103.2, and 99.1 x 10(-5) s/mm(2), respectively. The correlation between the ADC values and the degree of liver fibrosis was moderate (Spearmans test, rho = -0.36). There was a significant difference in ADC values between patients with nonfibrotic liver (F0) and those with cirrhotic liver (F4) (p = 0.008). The best cutoff ADC value to distinguish between these groups was 118 x 10(-5) s/mm(2). However, ADC values were not useful for differentiating viral hepatitis patients with F2 fibrosis or higher from those with a lower degree of fibrosis (area under the receiver operating characteristic curve [AUC] = 0.66) or for differentiating low-stage fibrosis in all patients from high-stage fibrosis in all patients (AUC = 0.54). CONCLUSION The ADCs in cirrhotic livers are significantly lower than those in nonfibrotic livers. However, ADC values measured using the current generation of scanners are not reliable enough to replace liver biopsy for staging hepatic fibrosis.

Automated bolus chase peripheral MR angiography: Initial practical experiences and future directions of this work-in-progress†

Bolus chase 3‐dimensional MR angiography (3D MRA) is a recent development that extends the effective field of view for arterial imaging from the typical single 40–50 cm to over 100 cm. This technique is well suited for imaging long vascular territories such as the lower extremity. Bolus chase peripheral 3D MRA is achieved with overlapping 3D gradient‐echo scans during the arterial transit of a single intravenous injection of gadolinium‐chelate contrast media. This technique can depict the arteries from the infrarenal aorta to the ankles in less than 2 minutes. The initial experiences with bolus chase peripheral MRA using an automated algorithm that controls both table translation and 3D data acquisition are described. Suggestions for future refinements to the technique are also discussed. J. Magn. Reson. Imaging 1999;10:376–388. Published 1999 Wiley‐Liss, Inc.

The value of diffusion-weighted imaging in characterizing focal liver masses.

RATIONALE AND OBJECTIVES To determine if focal liver masses could be differentiated as benign or malignant on the basis of diffusion-weighted imaging (DWI). METHODS AND MATERIALS A total of 104 patients with focal liver masses were scanned using 1.5 T magnetic resonance imaging (MRI). DWI was performed with b values of 0, 50, and 400 s/mm(2). Of these, 76 patients had lesions larger than 2 cm diameter, radiologic or pathologic characterization of the lesion, and diagnostic quality DWI. The apparent diffusion coefficient (ADC) of the largest liver lesion was measured. The liver masses were diagnosed on histology or had characteristic computed tomography/MRI findings and follow up of more than 6 months. The analyzed lesions were hemangioma (n = 17), cysts (n = 5), hepatocellular cancer (HCC) (n = 41), adenoma (n = 3), focal nodular hyperplasia (FNH) (n = 6), and metastases (n = 4). RESULTS The mean (standard deviation) ADC values (10(-5) mm(2)/second) of hemangiomas, cysts, FNH, and HCC were 156.8 (54.1), 190.2 (43.0), 130.1 (81.9), and 107.6 (32.7). The ADC of cysts and hemangiomas were significantly higher than that of other lesions (P = .0003, t-test). There was no significant difference between ADC values of solid, benign liver lesions (FNH, adenoma) and malignant lesions (HCC, metastases) (P = .62). CONCLUSION Solid liver lesions have a lower ADC than cysts and hemangiomas. However, there is no significant difference in ADC between solid benign and malignant lesions. DWI appears to have only minimal additional value over currently used MRI sequences in characterizing liver masses.

Local versus global features for content-based image retrieval

It is now recognized in many domains that content-based image retrieval (CBIR) from a database of images cannot be carried out by using completely automated approaches. One such domain is medical radiology for which the clinically useful information in an image typically consists of gray level variations in highly localized regions of the image. Currently, it is not possible to extract these regions by automatic image segmentation techniques. To address this problem, the authors have implemented a human-in-the-loop (a physician-in-the-loop, more specifically) approach in which the human delineates the pathology bearing regions (PBR) and a set of anatomical landmarks of the image at the time the image is entered into the database. From the regions thus marked, the approach applies low-level computer vision and image processing algorithms to extract features related to the variations of gray scale, texture, shape, etc. The extracted features create an index that characterizes the image. To form an image-based query the physician first marks the PERs. The system then extracts the relevant image features, computes the distance of the query image to all image indices an the database, and retrieves then most similar images. The approach is based on the assumption that medical image characterization must contain features local to the PERs. The focus of the paper is to assess the utility of localized versus global features for the domain of HRCT images of the lung, and to evaluate the systems sensitivity to physician subjectivity in delineating the PBRs.

Nonalcoholic fatty liver disease

OBJECTIVE The inflammatory subtype of nonalcoholic fatty liver disease, nonalcoholic steatohepatitis, is becoming one of the most important causes of chronic liver disease. In this article, we discuss the epidemiology, pathogenesis, and clinical and radiologic diagnosis of the subtypes of nonalcoholic fatty liver disease. CONCLUSION We discuss the current and evolving imaging tests in the evaluation of hepatic fatty content, inflammation, and fibrosis.

MR evaluation of liver iron overload

Children and young adults with hemolytic anemias requiring frequent transfusions develop increased liver iron content. We evaluated 15 chronically transfused children with sickle cell disease to determine whether spin-echo magnetic resonance (MR) imaging was useful in assessing the degree of iron overload. Quantitative MR parameters were correlated with liver biopsy iron determinations and serum ferritin levels. The best predictor of liver iron was the ratio of the intensities between the liver and paraspinal musculature on somewhat T1 weighted sequence (repetition time 0.5 s, echo time 28 ms). R2 = 0.58. Magnetic resonance was able to separate those patients with liver iron levels >100 μg/mg (intensity ratio ∼0.4). from those with levels <100 μg/mg (intensity ratios near I). However. MR was unable to quantitate liver iron in patients with values ranging from 100 to 400 μg/mg since similar intensity ratios were present in this range. Thus. MR provides a qualitative rather than quantitative assessment of liver iron overload.

Computed Tomography Demonstration of Lipomatous Metaplasia of the Left Ventricle Following Myocardial Infarction

Replacement of myocardium by fat, particularly of the right ventricle, is often diagnosed as arrhythmogenic right ventricular dysplasia. At autopsy, however, 68% of scars associated with chronic ischemic heart disease have shown fatty metaplasia in the scar. Four patients with a past history of previous myocardial infarctions and computed tomography demonstration of fatty change in left ventricular regions of hypokinesis and infarction are presented. It is proposed that these findings represent ischemic fatty metaplasia, an alternative etiology of fatty tissue replacing myocardium.