A Noninvasive Technology Using Artificial Intelligence to Measure the Diameter of Esophageal Varices Under Endoscopy

Abstract

Background: The diameter of esophageal varices (EVs) can not only predict variceal bleeding episodes but is also considered to be a significant factor in determining the endoscopic treatment of EVs. At present, visual observation is the most common method for estimating the diameter of EVs, the results of which may vary greatly between endoscopists. Materials and Methods: Herein, a noninvasive measurement technology, a virtual ruler (VR), was designed using artificial intelligence. The diameter and pressure of EVs in 7 patients were measured using VR and an esophageal varix manometer (EVM). Statistical methods, including the Bland-Altman Plot and Pearson correlation coefficient analysis, were used to compare the aforementioned 2 methods. Results: The results showed that the diameter of EVs measured using the aforementioned 2 methods did not differ. In addition, the time taken to measure the diameter of EVs using VR was 31 seconds (range, 25 to 44 s), significantly shorter compared with 159 seconds (range, 95 to 201 s) taken using an EVM (P< 0.01). Furthermore, the diameter of EVs measured using an EVM exhibited a high linear correlation with pressure. Conclusions: The current study demonstrated that VR was more accurate in measuring the diameter of EVs compared with EVMs while reducing unnecessary early intervention and the risk of complications. In terms of clinical risk and economic cost, this technology is hardly a burden. Overall, VR could be a useful software for the endoscopic detection and treatment of EVs in patients with liver cirrhosis.