Improved Capsule Endoscopy Using New Computer Vision Technologies / 대한내과학회지

ABSTRACT

Advanced computer vision technologies can be used to improve the performance of capsule endoscopy. Specifically, new technologies can be used to enhance the image quality of capsule endoscopes, which has previously been limited by mechanical constraints. The trajectory of a capsule endoscope, which can be estimated by reconstructing the captured images, helps to predict the location of lesions. Computer vision technology is utilized in a more practical way in a recently developed capsule endoscope comprising a stereo camera. Using this device, images can be interpreted in three dimensions, enabling more accurate measurements of size. The use of deep learning for image recognition is associated with an improved lesion detection rate and reduced reading time in clinicians. This evolving technology requires close cooperation between medical and IT professionals for optimized clinical use.