Chapter 13 - Magnetic capsule endoscopy: concept and application of artificial intelligence

Abstract

Magnetic capsule endoscopy (MCE) has aroused widely attention since Carpit initially put foward the concept of magnetic control in 2006. After efforts of technological optimization and development, current research has mainly focused on external magnetic field driving methods. Hand-held MCE, magnetic resonance imaging (MRI)-based MCE, and robotic MCE are the three major ways of achieving external active control of capsule gastroscopy. Besides the revolution in the magnetic field driving methods, the observing spectrum and parameters of MCE have extended a lot, including the emergence of a detachable string MCE (DS-MCE), a second-generation MCE (MCE-2), and so on. Artificial intelligence (AI), especially represented by the deep learning (DL) models with the use of the convolutional neural network, has emerged as an efficient and accurate automated image recognition method in various medical fields, showing remarkable performances in lesion identification or differentiation with high sensitivity and specificity. A few AI-based methodologies have been published in capsule localization and classifications of digestive lesions in MCE with satisfactory outcomes.