"4187 A COMPUTER-ASSISTED MODEL FOR TH.E ASSESSMENT OF ENDOSONGRAPHICALLY IDENTIFIED LYMPHADENOP~I'HY David E. Loren, Jill A. Patton, Gregory G. Ginsberg, Chandra M. Sehgal, Michael L. Koclnnan, Uulv of Pennsylvania Health System, Philadelphia, PA Background: Endoscopic ultrasound (EUS) is the most accurate means of tumor staging of esophageal carcinoma. While lymph node detection is high, discriminating between benign and malignant lymphadenopathy remains problematic. Features reported to favor malignant nodal disease include hypoechogenicity, spherical shape, sharply demarcated borders, size >lcm, image heterogenicity, and absence of a hilar echogenic stripe. Accordingly, an automated system that can yield an objective and reliable means of diagnosing lymphadenopathy is desirable. We applied software that allows for the delineation of whole-node contour and performs image analysis. Methods: Patients were identified who underwent EUS between 1996 and 1998 for evaluation of esophageal carcinoma. Inclusion in the study required that pathology of all nodes at a given anatomic site within a patient were either malignant or benign, and full pathologic, endosongraphic, and operative data were available. Patients were excluded if they had received neoadjuvant therapy. From this group a total of 19 malignant and 16 benign nodes were identified for characterization. Utilizing a program developed in conjunction with the Ultrasound Research Laboratory, endosonographic still images were digitized, nodal borders were traced, and measurements were made of echogenicity, whole-node heterogeneity, and regional variability. Echogenicity was determined by mean brightness of pixels comprising the image. Whole-node heterogeneity was evaluated by assessing the variability (standard deviation) ofpixel distribution of the entire node. Regional variability was assessed by automated sectioning of the nodal image into five concentric slices based on the outlined border. The variability of brightness values of each slice was calculated as the measure of regional variability. Results: 1. Malignant nodes were significantly hypcechoic, (mean=lO4) relative to benign nodes (mean=125), p<0.04. 2. Total node heterogeneity as measured by the standard deviation of pixel distribution was significantly greater for malignant nodes than benign nodes p<0.008. 3. There was a non-significant trend for greater regional variability for benign nodes p=0.09. Conclusion: This preliminary study demonstrates the feasibility of an automated system to designate malignant lymphadenopathy. Further development of this system may allow for selective stratification of patients for stage-specific therapy. "4188 IS HIGH-FREQUENCY IYLTRASOUND PROBES NECESSARY IN THE INVASION DEPTH DIAGNOSIS OF EARLY COLORECTAL CARCINOMAS? Mikihiro Fujiya, Yusuke Saitoh, Jiro Watari, Kaori Fujiya, Atsuo Maemoto, Fumika Orii, Tokiyoshi Ayabe, Toshifumi Ashida, Takashi Obara, Yutaka Kohgo, Asahikawa Medical Coil, Asahikawa Japan Background: Endoscopic resection (ER) is indicative for intramucosal (m) or focally extended submucosal (sml) cancers because lymph node metastasis in s m l cancers xvere extremely rare but those in moderately (sin2) or massively extended (sin3) submucesai cancers were considered to be -10%. It is important to differentiate m-sml cancers from sin2-3 cancers. High-frequency ultrasound probes (HFUP) has become available as the modality for selecting the appropriate therapeutic option in early eolorectai cancers (m and sml-3 cancers). Aim: The aim of this prospective study is to elucidate the efficacy of HFUP in the invasion depth diagnosis for the choice of therapy in early eolorectal cancers. Material and Methods: 122 early colorectal cancers which were diagnosed and treated in our Hospital were enrolled for this prospective study. Colonoscopic diagnosis was made firstly, then, HFUP diagnosis was made independently without other information before treatment. The colonoscopic and the HFUP diagnosis were made according to our criteria (Saitoh et al. GIE 1998, 48:362-70, GIE 1996, 44:34-9). We prospectively compared diagnostic values for the invasion depth diagnosis made by HFUP and those by colonoscepy. Results:l. Those early colorectal cancers were categorized into 92 m-sml cancers and 30 sin2-3 cancers. 2. There was no significant difference (p=0.382) in overall accuracy for the invasion depth diagnosis between colonoscepy (100/122, 82%) and HFUP (105/122, 86%). 3. In the diagnosis of sin2-3 cancers, colonoscopic diagnosis had 67% sensitivity, 87% specificity and 63% positive predictive value, and HFUP had 86%, 87% and 67%, respectively. 4. The causes of 17 misdiagnosed cases by HFUP were ultrasound attenuation in 7 cases, difficulty in proper scan in 3 cases and submucosal fibrosis or lymphoid follicle in 5 cases. 5. HFUP had correct diagnosis in 9 of 22 misdiaguosed cases by celonoscopy. Of those, 6 cases were fiat & depressed cancers macroscopically and/or 6 cases were sin2-3 cancers in the invasion depth. Conclusion: There is no significant difference in overall accuracy for the invasion depth diagnosis between colonoscepy and HFUP. However the lesion is suspected of a submucosal cancer by colonoscopy in flat & depressed type need additional HFUP for obtaining appropriate therapeutic strategy.
AB164
GASTROINTESTINAL ENDOSCOPY
"4189 MALT GASTRIC LYMPHOMA: DIAGNOSIS AND 30 MONTHS FOLLOW UP AFTER HELICOBACTER PYLORI TREATMENT USING ENDOSCOPIC ULTRASONOGRAPHY. Luis C. Sabbagh, Benedicte Velasco, Clin Reina Sofia, Bogota Colombia Introduction. Low grade MALT gastric lymphoma (mucosa associated lymphoid tissue)is directly related to Helicebacter pylori infection. Endoscopic Ultrasonography (EUS) is the gold standard to evaluate the gastric wall thickness, tumor extension and peri gastric node involvement. The objective of this study was to prospectively evaluate the utility of EUS in the follow up of patients with MALT Lymphoma stage 1A 30 months after Helicebacter pylori treatment. Methods The inclusion criteria were patients with histology and immuno histochemistry diagnosis of low grade MALT gastric lymphoma with superficial involvement (no deeper than submucosa layer and without node involvement). EUS was performed before and on the 3rd, 6th, 9th, 12th, 18th, 24 th and 30 month after antibiotic treatment with ciaritromicin, amoxaeilin and lansoprazol for Helicobacter pylori eradication. The main outcome variable was defined as the measurement of the tumor involvement of the gastric wall (in millimeters, normal value of 3 ram) by EUS on the specified times. At the same time, biopsy samples were taken. The EUM 20 Olympus Endosonography unit was used. Results. The study period of time was 30 months; during this time, 23 patients were included. The median of the gastric mucesa thickness before treatment was 8 mm (range 3 - 10 ram). Three patients remained with the same gastric mucosa thickness and positive biopsy for the neoplasm. There was a significant reduction of the gastric mucosa thickness after the sixth month of treatment in 20 patients (86.9%), with a median of 4 ram. The tumor was not evident at the biopsy specimen in this group of patients and the eradication of the Helicobacter pylori was demonstrated. There was a good correlation between the EUS evaluation and the histological findings for tumor evidence and for Helicobacter pylori infection. After 30 months of follow up we didn't find relapse in the group ~vith good initial response. Conclusions: Our findings showed a good correlation between H pylori eradication and gastric mucosa thickness reduction in patient with MALT gastric lymphoma (1A stage) during 30 months follow up. EUS was a useful tool for the diagnosis and follow up in this group of patients. "4190 T H E P R E D I C T I V E VALUE OF A NEGATIVE ENDOSCOPIC ULTRASOUND IN THE ASSESSMENT OF SUSPECTED PANCREATIC CARCINOMA Alphonso Brown, Nuzhat A. Abroad, James D. Lewis, Michael L. Kochman, Sandford J. Schwartz, Mark Weiner, Gregory G. Ginsberg, Univ of Pennsylvania, Philadelphia, PA Background: CT imaging studies performed in patients with vague abdominal symptoms commonly raises suspicions of pancreatic neoplasms. Although endoscopic ultrasound (EUS) has been used to evaluate patients with suspected pancreatic lesions, it's ability to predict the true absence of pancreatic malignancy is not known. Purpose: To define the ability of EUS to correctly classify the absence ofmaliguancy in patients who are referred with a suspected pancreatic mass. Methods: We conducted a retrospective cohort study of all patients who underwent EUS evaluation of the pancreas from December 1993 - December 1997 in whom a CT scan raised suspicion of a tumor of the pancreas. An EUS exam was considered negative if no discrete hypoechoic mass or cystic lesion was identified. Clinical follow-up was obtained by interviewing referring physicians, reviewing medical records, and searching the Social Security Death Index (SSDI). An EUS exam was considered a false negative if a patient was diagnosed with pancreatic cancer following a negative EUS. Patients with less than 1 year of follow-up data were excluded (n=20) unless pancreatic cancer was diagnosed within the first year of follow-up. Results: 292 patients underwent EUS of the pancreas for a suspected pancreatic mass lesion. Of these, 93 (32%) had negative EUS exams. 73 of the 93 patients met our inclusion criteria (median follow-up 4 years, range 1-7 years). Of these 8 had EUS features of chronic panereatitis (CP), including increased echogenicity, heterogeneity, increased internal reflectors and ductal changes. Three of the 73 patients were subsequently diagnosed with pancreatic cancer. Each of these 3 patients had EUS features of CP. Based on these 73 patients, the NPV of EUS was 96% (95% CI 88% - 99%). Of the 20 excluded patients in whom complete clinical follow-up was not available, search of the SSDI indicated that 4 of the 20 patients had died. The median time from EUS to search of SSDI in these patients was 4 years (range 3-7 years). Assuming that each of the 4 patients had died of pancreatic cancer and no other patients developed pancreatic cancer, the NPV of EUS was 92% (95% CI 85% - 97%). Conclusion: EUS evaluation of the pancreas in patients with suspected pancreatic cancer has a high NPV. The predictive value is lower in patients with EUS features of chronic pancreatitis.
VOLUME 53, NO. 5, 2001