What we should expect from artificial intelligence in video capsule endoscopy

We read with interest the manuscript by Ding et al. published in Endoscopy [1]. The authors aimed to develop an artificial intelligence (AI) model capable of diagnosing eight categories of abnormalities and to compare its diagnostic performance with that of doctors of different experience levels. The results of the study are relevant for several reasons. First, the AI software systems available to date for small-bowel capsule endoscopy (SBCE) have been designed mainly for detection and discriminating between normal and abnormal images. The Ding et al. model is the first software to be developed and validated for lesion characterization in this setting. This represents a major development, as it further influences diagnostic outcomes by selecting specific pathological images instead of the wide spectrum of “abnormal images,” including irrelevant findings. Second, evidence available in the literature mainly comes from proof-of-concept studies that used altered images and/or video segments [2] [3]. Therefore, published results can be easily misinterpreted without a detailed understanding of the pitfalls, and the real-world AI performance might not be replicated. In the validation phase of the Ding et al. study, full videos were analyzed. This represents a relevant strength, as trials using unaltered images and including comparison with standard care represent a clinical priority in confirmation of the reported expert level performance of AI software. Third, the study provides a picture of the potential role of AI in SBCE training, showing that trainees can easily achieve the performance of experts with the assistance of AI [4]. AI is revolutionizing medicine, and digestive endoscopy is one of the medical fields in which it is spreading fastest. In SBCE, detection, characterization, training optimization, and reduction of reading time represent only the initial steps of AI application; in the future we expect the role of AI to expand further to include, but not limited to, size measurement, preparation quality evaluation, localization, and evaluation of quality measures.

Publication History

Article published online:
27 July 2023

• References

• 1 Ding Z, Shi H, Zhang H. et al. Artificial intelligence-based diagnosis of abnormalities in small-bowel capsule endoscopy. Endoscopy 2023; 55: 44-51
• 2 Qin K, Li J, Fang Y. et al. Convolution neural network for the diagnosis of wireless capsule endoscopy: a systematic review and meta-analysis. Surg Endosc 2022; 36: 16-31
• 3 Piccirelli S, Milluzzo SM, Bizzotto A. et al. Small bowel capsule endoscopy and artificial intelligence: first or second reader?. Best Pract Res Clin Gastroenterol 2021; DOI: 10.1016/j.bpg.2021.101742.
• 4 Piccirelli S, Bizzotto A, Pesatori EV. et al. PO951 Role of artificial intelligence in small bowel capsule endoscopy training. United European Gastroenterology J 2022; 10 (Suppl. 08) 1046