Subscribe to RSS
DOI: 10.1055/s-0040-1704291
A DEEP LEARNING-BASED SYSTEM FOR IDENTIFYING DIFFERENTIATION STATUS AND DELINEATING MARGINS OF EARLY GASTRIC CANCER IN NARROW-BAND IMAGING ENDOSCOPY
Publication History
Publication Date:
23 April 2020 (online)
Aims Accurate differentiation status diagnosis of early gastric cancer and cancer margin delineation are critical for the treatment strategy and achieving endoscopic curative resection. The aim of this study was to train a real-time system to accurately identifying differentiation status and delineating margins of EGC in ME-NBI endoscopy.
Methods 2217 magnifying narrow-band images (ME-NBI) from 145 EGC patients and 882 images from 58 EGC patients were used to train and test the convolutional neural network (CNN) 1 for identifying EGC differentiation status. 256 images from 67 EGC patients and 69 images from 31 EGC patients were used to train and test CNN2 for delineating EGC margins.
Results The system correctly predicted differentiation status of EGCs with an accuracy of 68.97%, on par with the performance of the five experts (69.66±13.33%, p = 0.91). For delineating EGC margins, the system achieved an accuracy of 93.75% in differentiated EGC and 97.37% in undifferentiated EGC under an overlap ratio of 0.50.
Conclusions We developed a deep learning-based system for accurately identifying differentiation status and delineating margins of EGC in ME-NBI endoscopy. This system achieved a performance comparable with experts, and was successfully tested in real EGC videos.
#