Elastic-Scattering Spectroscopy (ESS) Forceps for Real-Time Classification of Diminutive Colorectal Polyps: Presidential Poster

Abstract

Purpose: There has been considerable interest in optical technologies that permit “real-time histology” of diminutive colorectal polyps to decrease biopsy risks and pathology costs, as recently highlighted in the ASGE PIVI guidelines [Rex, GIE, 2011]. Our aim was to evaluate the performance of ESS, mediated through optically-integrated disposable forceps, as a simple, low cost, automated platform for in situ classification of diminutive polyps during colonoscopy. Methods: In an IRB-approved study, patients undergoing screening/surveillance colonoscopy at VA Boston were recruited. The system consists of a standard disposable-type biopsy forceps with optical fibers integrated between the jaws, a control system incorporating a xenon lamp, spectrometer, and CPU with custom software. The ESS signal derives from the wavelength dependent scattering and absorption properties of the measured tissue, over the white light spectrum, and is sensitive to cellular and subcellular architecture, including nuclear and organellar size and density, as well as to hemoglobin content/vascularity. The optical forceps were used whenever endoscopic tissue sampling was indicated per standards of care. The forceps were placed in gentle contact with tissue to be biopsied, measurements taken in less than 1 sec, and the jaws closed to permit precise co-registration of the biopsy and volume of tissue optically measured. Biopsies were interpreted independently by three GI pathologists and the consensus classification was correlated to the optical measurements. A “black box” approach was used to classify the spectra, utilizing high-dimensional machine learning approaches. Results: A total of 122 non-adenomas and 34 adenomas (≤5mm) from 69 patients were interrogated with ESS before biopsy. Retrospective leave-one-patient-out analysis resulted in a Se=95%, Sp=89%, and NPV=99% with an overall accuracy of 90% from high-confidence readings. A sub-analysis performed on ≤5mm recto-sigmoid polyps (82 non-adenomas, 7 adenomas) resulted in a Se=100%, Sp=89%, and NPV=100% with an accuracy of 90%. Conclusion: ESS is a potentially practical, low cost, familiar platform for in situ classification of diminutive polyps requiring virtually no learning curve. The accuracy, and NPV on rectosigmoid polyps will likely meet the PIVI guidelines of ≥90% agreement with histopathology in predicting surveillance intervals for a resect and discard approach, and of NPV≥90% for leaving behind diminutive recto-sigmoid polyps. Prospective validation on larger cohorts is planned to establish clinical efficacy per the colon PIVI guidelines.