S3244 EUS Elastography in Prediction of Lymph Node Metastases in Suspected GI Cancer: A Pilot Study

Abstract

INTRODUCTION: Endoscopic ultrasound (EUS) and EUS-guided fine-needle aspiration (EUS-FNA) are the most accurate techniques for local staging of malignant gastrointestinal (GI) cancers. Four features are used to predict metastatic lymph nodes (LNs) with 80% accuracy if all are present: hypoechogenicity, well-defined margins, round shape, and size >1 cm. Despite near 100% specificity, the challenges of obtaining a sample limits the use of FNA. EUS elastography is a real-time technology (EUS-RTE) with the potential to differentiate malignant and benign tissue via color histogram. Investigators have speculated that the subjectivity of color interpretation leads to variable results. Image J is a processing software that can be used to help minimize bias. Our goal is to establish a quantitative EUS criterion for assessment of LNs. METHODS: At least 3 EUS-RTE images of each LN were obtained. The endoscopist recorded his assessment of the LN based on the four criteria described above (“endoscopist’s impression”). If at least 3 of the 4 criteria were present, the LN was considered malignant. FNA was performed if feasible. All images were from a linear echoendoscope. Images without adequate transducer pressure as measured by the strain graph display or cytology were excluded. A stylus was used to outline the entire LN in B-mode, which was considered the region of interest (ROI). This outline was superimposed on the corresponding color histogram for processing. The number of red, green and blue pixels and total area of each ROI were quantified using Image J. RESULTS: Twenty-one LNs from 19 patients were included in the final analysis. Ten LNs were malignant (45%). Most procedures were performed for suspected esophageal or pancreatic cancer (14/19). A comparison of the endoscopist’s impression, blue pixels as a predictor of malignancy and red pixels as a predictor of benign LN is depicted in Figure 1. Sensitivity and specificity of the endoscopist’s impression was 90% and 67% respectively; accuracy was 80%. The endoscopist’s impression performed the best (95% CI, 0.63-0.98), but the difference was not statistically significant. CONCLUSION: Our study was limited by a small sample size and selection of the ROI may be another source of subjectivity. While Image J did not outperform current EUS criteria, firm conclusions cannot be drawn. Further study with a larger cohort is necessary to refine our interpretation of elastography and determine if post-procedure processing of elastography images is a useful adjunct.Figure 1