2023-11-28 15:57:02
Amsterdam UMC and Eindhoven University of Technology researchers have developed an artificial intelligence (AI)-based model to support endoscopists in recognizing early neoplasia in patients with Barrett’s esophagus. The study, published in The Lancet Digital Health, found that the AI model significantly improved sensitivity in detecting neoplastic changes of the mucous membrane, which can often be subtle and difficult to recognize.
Barrett’s esophagus patients have a slightly less than 5 percent increased risk of esophageal carcinoma and are checked endoscopically regularly. However, detecting neoplastic changes is challenging, and missing a lesion can have significant consequences for the prognosis. Surveillance intervals for these patients can be as long as five years.
The computer model was tested by 112 endoscopists from six different countries. The results showed that sensitivity, based on photos, increased from 74 to 88 percent with the use of the AI support. Similarly, when rating videos, sensitivity increased from 67 percent to 79 percent. The researchers noted that the specificity remained more or less the same with the AI support.
To train the computer model, researchers used a general database called ImageNet, followed by training the model with GastroNet, a dataset that includes approximately fourteen thousand photos and videos of neoplastic and non-dysplastic Barrett’s esophagus. The study also highlighted a weakness in the training as the model would need to be fed with images from general endoscopic centers, which are often less sharp.
The development of this AI-based model represents a significant advancement in the field of gastroenterology and could potentially improve the detection and treatment of neoplastic changes in patients with Barrett’s esophagus. Further research and development are needed to refine and enhance the model’s capabilities.]
#model #endoscopic #Barretts #esophagus