A Deep Learning-based System for Identifying Differentiation Status and Delineating Margins of Early Gastric Cancer in Magnifying Narrow-band Imaging Endoscopy.

Researchers

Journal

Modalities

Models

Abstract

Accurate identification of early gastric cancer (EGC) differentiation status and margins is critical for determining the surgical strategy and achieving curative resection in EGC patients. The aim of this study was to develop a real-time system for accurately identifying differentiation status and delineating margins of EGC in Magnifying Narrow-band Imaging (ME-NBI) endoscopy.
2217 images from 145 EGC patients and 1870 images from 139 EGC patients were retrospectively collected to train and test the convolutional neural network (CNN) 1 for identifying EGC differentiation status. 882 images from 58 EGC patients were used to compare the performance of CNN1 with that of experts. 928 images from 132 EGC patients and 742 images from 87 EGC patients were used to train and test CNN2 for delineating EGC margins.
The system correctly predicted differentiation status of EGCs with an accuracy of 83.3% (95%CI: [81.5%-84.9%]) in testing dataset. In the man-machine contest, CNN1 performed significantly better compared to the five experts [86.2% (95%CI: [75.1%-92.8%]) vs. 69.66 (95%CI: [64.1%-74.7%])). For delineating EGC margins, the system achieved an accuracy of 82.7% (95%CI: [78.6%-86.1%]) in differentiated EGC and 88.1% (95%CI: [84.2%-91.1%]) in undifferentiated EGC under an overlap ratio of 0.80. In unprocessed EGD videos, the system achieved real-time EGC differentiation status diagnosis and EGC margin delineation in ME-NBI endoscopy.
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 superior performance when compared with experts, and was successfully tested in real EGC videos.
Thieme. All rights reserved.