Post written by Yusuke Horiuchi, MD, PhD, from the Department of Gastroenterology, Cancer Institute Hospital, Tokyo, Japan.
The performance of magnifying endoscopy with narrow-band imaging (ME-NBI) using a computer-aided diagnosis (CAD) system in diagnosing early gastric cancer (EGC) is unclear. Here, the focus of our study is to clarify the differences in the diagnostic performance between expert endoscopists and CAD system using ME-NBI. For accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV), comparisons were made between the CAD system and 11 experts who were skilled in diagnosing EGC using ME-NBI.
ME-NBI is a form of advanced endoscopy whose performance in diagnosing EGC has been reported to be better than that of conventional endoscopy. However, diagnosing EGC with ME-NBI requires considerable skills. Additionally, in cases of severe inflammatory cell infiltration, more advanced diagnostic techniques are required, and diagnosing EGC with ME-NBI is challenging.
In recent years, artificial intelligence has been developed and applied in the medical field. The construction of a system for classifying learned images in convolutional neural network (CNN) has definitely improved the image recognition capabilities of CAD systems. Therefore, we believe that a CAD system may be useful in the diagnosis of EGC using still ME-NBI images.
The CAD system demonstrated an area under the curve of 0.8684. The accuracy, sensitivity, specificity, PPV, and NPV were 85.1% (95% confidence interval [95% CI]: 79.0–89.6), 87.4% (95% CI: 78.8–92.8), 82.8% (95% CI: 73.5–89.3), 83.5% (95% CI: 74.6–89.7), and 86.7% (95% CI: 77.8–92.4), respectively. The CAD system was significantly more accurate than 2 experts, significantly less accurate than 1 expert, and not significantly different from the remaining 8 experts.
The overall performance of the CAD system using ME-NBI videos in diagnosing EGC was considered high and was equivalent to or better than that of several experts. The CAD system may prove useful in the diagnosis of EGC in clinical practice.
This study had several limitations. The videos used for this study were retrospectively collected from a single center and were not actual real-time endoscopy videos in which the CAD system established the diagnoses. In addition, because only ESD cases were included, there are cases in which EGC was overlooked and ESD was not performed. Therefore, this false negative may not be a real false negative.
We plan to propose a prospective study that clarifies the additional effect of the CAD system for experts by comparing the group in which the diagnosis by the CAD system is used together with the diagnosis by experts and the group in which the diagnosis by experts alone is used in real-time endoscopy.
Read the full article online.
The information presented in Endoscopedia reflects the opinions of the authors and does not represent the position of the American Society for Gastrointestinal Endoscopy (ASGE). ASGE expressly disclaims any warranties or guarantees, expressed or implied, and is not liable for damages of any kind in connection with the material, information, or procedures set forth.