AI-based Diagnostic Tool Shows Promise in Upper GI Cancer

Article

A new study found artificial intelligence-based diagnostics could help with better diagnoses for upper gastrointestinal cancers.

A new system of artificial intelligence-based diagnostics using endoscopic imaging was highly accurate in diagnosing upper gastrointestinal cancers. The system, termed the Gastrointestinal Artificial Intelligence Diagnostic System (GRAIDS), could help community hospitals in improving gastrointestinal cancer diagnoses, according to the results of a study published in Lancet Oncology.          

“Most upper gastrointestinal cancers are diagnosed at advanced stages because their signs and symptoms tend to be latent and non-specific, leading to an overall poor prognosis, but if detected early, 5-year survival can exceed 90%,” wrote study authors led by Huiyan Luo, MD, of Sun Yat-sen University Cancer Center in Guangzhou, China. “Novel endoscopic strategies and techniques have been developed for these cancers, but they may not be effective in all areas. The risk of missing suspicious upper gastrointestinal cancers in endoscopy examinations might still be high in hospitals with low patient volume, in less developed or remote regions, and even in countries where many endoscopies are performed.”

 The researchers used a total of 1,036,496 endoscopy images from 84,424 individuals at 6 hospitals in China to develop and test the new GRAIDS method. They used only images from patients with histologically confirmed esophageal or gastric cancer and randomized individuals to training, intrinsic verification, and internal validation sets

 The GRAIDS algorithm was based on Google’s DeepLab V3+. Inputting the endoscopy images resulted in an output of whether or not the image contained a tumor, and then a segmentation of the tumor region of the image.

After development of the algorithm, GRAIDS achieved a diagnostic accuracy in identifying upper GI cancers in the internal validation set of 0.955. In 5 external validation sets, its accuracy ranged from 0.915 to 0.977. The algorithm’s diagnostic sensitivity was similar to that of an expert endoscopist, at 0.942 compared with 0.945 (P= 0.692). It was superior to that of a “competent” endoscopist (0.858; P< 0.0001) and a trainee (0.722; P< 0.0001).

The positive predictive value for GRAIDS was 0.814, compared with 0.932 for an expert endoscopist, 0.974 for a competent endoscopist, and 0.824 for a trainee. The negative predictive value for GRAIDS was 0.978, compared with 0.980, 0.951, 0.904 for the 3 endoscopists, respectively.

“To the best of our knowledge, this is the largest study in the field of artificial intelligence-guided cancer detection based on upper gastrointestinal endoscopic images worldwide,” the authors wrote. “GRAIDS could support non-expert endoscopists by improving their diagnostic accuracy to a level similar to that of experts.”

References:

Luo, H., Xu, G., Li, C., He, L., Luo, L., Wang, Z., Jing, B., Deng, Y., Jin, Y., Li, Y., Li, B., Tan, W., He, C., Seeruttun, S., Wu, Q., Huang, J., Huang, D., Chen, B., Lin, S., Chen, Q., Yuan, C., Chen, H., Pu, H., Zhou, F., He, Y. and Xu, R. (2019). Real-time artificial intelligence for detection of upper gastrointestinal cancer by endoscopy: a multicentre, case-control, diagnostic study. [online] Available at: https://www.thelancet.com/journals/lanonc/article/PIIS1470-2045(19)30637-0/fulltext [Accessed 9 Oct. 2019].

Recent Videos
Immunotherapy may be an “elegant” method of managing colorectal cancer, says Gregory Charak, MD.
Administering neoadjuvant therapy to patients with colorectal cancer may help surgical oncologists attain a negative-margin resection.
Increasing screening for younger individuals who are at risk of colorectal cancer may help mitigate the rising early incidence of this disease.
Laparoscopy may reduce the degree of pain or length of hospital stay compared with open surgery for patients with colorectal cancer.
Rahul Gosain, MD; Sam Klempner, MD; and Rohit Gosain, MD, presenting slides
Rahul Gosain, MD; Sam Klempner, MD; and Rohit Gosain, MD, presenting slides
Rahul Gosain, MD; Sam Klempner, MD; and Rohit Gosain, MD, presenting slides
Rahul Gosain, MD; Sam Klempner, MD; and Rohit Gosain, MD, presenting slides
Rahul Gosain, MD; Sam Klempner, MD; and Rohit Gosain, MD, presenting slides
Tailoring neoadjuvant therapy regimens for patients with mismatch repair deficient gastroesophageal cancer represents a future step in terms of research.
Related Content