This study provided 936 images of esophageal cancer endoscopy as a training image, including 498 white light endoscopes (WLI) and 438 narrow-band imaging endoscopes (NBI) images. According to the esophageal cancerization process, it is divided into four types: metaplasia (Dysplasia), metaplasia and esophageal cancer (Dysplasia-ECA), and esophageal cancer (ECA). A Single Shot Multibox Detector (SSD) was constructed by Convolutional Neural Network (CNN), and 264 test images were prepared to evaluate the accuracy of the model diagnosis. We developed a system called DNN-CAD to identify neoplastic or hyperplastic colorectal polyps less than 5 mm. The system classified polyps with a PPV of 89.6%, and a NPV of 91.5%, and in a shorter time than endoscopists. This deep-learning model has potential for not only endoscopic image recognition but for other forms of medical image analysis, including sonography, computed tomography, and magnetic resonance images. The construction of an SSD system for detecting esophageal cancer can analyze stored endoscopic images with high sensitivity in a short time, but more training can improve the accuracy of diagnosis. The system can facilitate early detection in practice and thus have a better diagnosis in the future. |
ACCESS THE FULL ARTICLE
No SPIE Account? Create one
Cancer
Diagnostics
Endoscopy
Esophagus
Artificial intelligence
Convolutional neural networks
Sensors