PURPOSE: The responses of polyps to light essentially determine the diagnostic capability of an endoscopy system in differentiating adenomas from hyperplastic polyps. Compared with white light colonoscopy (WLC), narrow-band imaging (NBI) is expected to improve the diagnostic capability. The diagnostic capabilities of WLC and NBI are evaluated and compared based on the polyp responses. METHODS: The following WLC and NBI images were retrospectively reviewed and categorized: 195 images and polyps (89 WLC, 106 NBI) with the best visual quality were categorized in the best image group (BG), and 484 images of 242 polyps (both WLC and NBI) were categorized in the paired image group (PG). For each reflection of light used for WLC or NBI, the polyp responses were objectively expressed as reflection features. The reflection features were then used to establish a classification model for identifying adenomas. The diagnostic capability of reflection feature or classification model was measured by the area under the receiver operating characteristic curve (AUC). RESULTS: In both image groups, the diverse and heterogeneous features of the polyp responses enabled accurate identification of adenomas, regardless of the light source used for WLC and NBI. For differential diagnosis of adenomas and hyperplastic polyps, the WLC and NBI did not significantly differ in BG (AUC, 0.905 and 0.922, respectively; P = 0.690) or in PG (AUC, 0.782 and 0. 769, respectively; P = 0.755). CONCLUSIONS: Using WLC and NBI as classification models is effective in differential diagnosis of colorectal polyps and exhibited similar capabilities.
PURPOSE: The responses of polyps to light essentially determine the diagnostic capability of an endoscopy system in differentiating adenomas from hyperplastic polyps. Compared with white light colonoscopy (WLC), narrow-band imaging (NBI) is expected to improve the diagnostic capability. The diagnostic capabilities of WLC and NBI are evaluated and compared based on the polyp responses. METHODS: The following WLC and NBI images were retrospectively reviewed and categorized: 195 images and polyps (89 WLC, 106 NBI) with the best visual quality were categorized in the best image group (BG), and 484 images of 242 polyps (both WLC and NBI) were categorized in the paired image group (PG). For each reflection of light used for WLC or NBI, the polyp responses were objectively expressed as reflection features. The reflection features were then used to establish a classification model for identifying adenomas. The diagnostic capability of reflection feature or classification model was measured by the area under the receiver operating characteristic curve (AUC). RESULTS: In both image groups, the diverse and heterogeneous features of the polyp responses enabled accurate identification of adenomas, regardless of the light source used for WLC and NBI. For differential diagnosis of adenomas and hyperplastic polyps, the WLC and NBI did not significantly differ in BG (AUC, 0.905 and 0.922, respectively; P = 0.690) or in PG (AUC, 0.782 and 0. 769, respectively; P = 0.755). CONCLUSIONS: Using WLC and NBI as classification models is effective in differential diagnosis of colorectal polyps and exhibited similar capabilities.
Authors: Louis Michel Wong Kee Song; Douglas G Adler; Jason D Conway; David L Diehl; Francis A Farraye; Sergey V Kantsevoy; Richard Kwon; Petar Mamula; Betsy Rodriguez; Raj J Shah; William M Tierney Journal: Gastrointest Endosc Date: 2008-04 Impact factor: 9.427