Post written by Mohamed Hussein, MRCP, from the Department of Gastroenterology, Guy’s and St Thomas’ NHS Foundation Trust, London, United Kingdom.
We aimed to develop a computer-aided characterization system that could support the diagnosis of dysplasia in Barrett’s esophagus on magnification endoscopy. We also strived to assess the speeds of the networks in making this diagnosis.
Despite advances in endoscopic imaging in the last 2 decades, there is still a significant rate of missed early cancer with Barrett’s esophagus. Curative endoscopic therapy can be offered if the diagnosis is picked up early, which is where the role of artificial intelligence (AI) will be important to try and offset some of these missed esophageal cancer rates.
Magnification imaging is useful in improving the detection and diagnosis of early cancer in Barrett’s esophagus. Because of some of the complex criteria for diagnosis on this imaging modality, there is a low uptake of this useful imaging technique in many hospitals. This is why we felt it would be of benefit to automate this using an AI system.
The main benefits of an AI system to characterize early cancer on magnification imaging would be to increase the uptake of this imaging modality in more centers, therefore improving early detection and diagnosis rates of early cancer.
However, it also would be beneficial for experts in delineating resection margins, ensuring a complete R0 resection during an EMR or endoscopic submucosal dissection.
On 49,726 magnification video frames, the AI system achieved a sensitivity of 92% in characterizing dysplasia. On high-quality images, the network achieved a sensitivity of 94% and specificity of 86%.
The mean assessment speed per frame was 0.0135 seconds (SD ±0.006). The system would be able to potentially work in real time to support the decision of endoscopists.
Such an AI system also would potentially reduce unnecessary biopsies and allow for more accurate endoscopic resections.
The next steps now are for a multicenter randomized control trial of AI systems for both detection and characterization of early cancer in Barrett’s esophagus that would create the path for the introduction of these AI systems in the clinical setting.
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.