Deep Learning Could Improve Esophageal Cancer Screening
A groundbreaking study has revealed that deep learning technology has the potential to revolutionize the early detection of esophageal cancer. This advanced system significantly enhances the ability of clinicians to identify high-risk esophageal lesions, including cancer and precancerous cells, during routine endoscopy procedures.
Researchers conducted a randomized, controlled trial, which demonstrated that the artificial intelligence system nearly doubled the detection rate of high-risk esophageal lesions when compared to unassisted endoscopy. The study, published in Science Translational Medicine, reported that the deep learning system identified one additional positive high-risk case per 111 patients screened.
Dr. Shao-Wei Li and his team at Taizhou Hospital of Zhejiang Province in China developed the real-time detection system, known as the ENDOANGEL-esophageal lesion detection system (ELD). This system is based on deep convolutional neural networks (CNN) and was trained using a vast dataset of over 190,000 esophagogastroscopic images.
During the trial, over 3,000 patients aged 50 years and above were randomly assigned to receive either CNN-assisted endoscopy or traditional endoscopy without AI assistance. The results showed a significantly higher detection rate of high-risk esophageal lesions in the group that received deep learning assistance.
The ENDOANGEL-ELD system demonstrated impressive sensitivity, specificity, and accuracy rates of 89.7%, 98.5%, and 98.2%, respectively, for detecting high-risk esophageal lesions. Importantly, no adverse events were reported during the study.
While the system exhibited three false negatives, the researchers emphasized the importance of continuous learning and improvement through feedback mechanisms. They plan to expand the system’s training data to encompass diverse clinical scenarios, allowing the AI model to identify a broader range of lesions accurately.
In conclusion, the ENDOANGEL-ELD system has proven to be effective and safe in assisting endoscopists in diagnosing high-risk esophageal lesions. This technological advancement has the potential to enhance the screening and detection rates of esophageal cancer, leading to earlier diagnosis and improved patient outcomes.
