Generation of magnifying endoscopic images of gastric neoplasms based on an all-in-focus algorithm

Background and Aim: Magnifying endoscopy is useful for diagnosis of early gastrointestinal neoplasms by visualizing microvascular (MV) and microsurface (MS) structures of the mucosa when combined with image-enhanced endoscopy. However, precise control of the endoscope is needed because the depth of focus is narrow and the target may move. These problems may be overcome by the all-in-focus (AIF) technique, which was developed in the engineering field. The aim of the study was to evaluate magnifying endoscopic image with AIF algorithm. Methods: Twenty gastric neoplasms were examined. Images were acquired at 80× magnification and converted to endoscopic images with an AIF algorithm (EI-AIF). The focus area and MV and MS patterns in the original image and the EI-AIF were compared on a 5-point Likert scale, where 5 indicates that the EI-AIF was superior. Intraclass correlation coefficients (ICCs) were used to assess the inter-evaluator reliability. An image quality measurement value was calculated for each image as an indicator of the degree of focus. Results: The scores for focus area, MV, and MS were 4.78 ± 0.45 (ICC = 0.63), 4.12 ± 0.76 (ICC = 0.70), and 4.72 ± 0.52 (ICC = 0.45), respectively, with the EI-AIF significantly superior for all three items (P < 0.05 by Student's t-test). ICCs for the focus area and MV were > 0.60, indicating strong inter-evaluator reliability. Image quality measurement was higher for the EI-AIF compared with the original image in every case. Conclusions: Endoscopic observation with AIF algorithm gives a better image quality that allows easier evaluation of MV and MS patterns. This technique may resolve the difficulties with magnifying endoscopic observation.