Spotting the blind spots
An AI-based system such as ENDOANGEL could overcome some of the natural weaknesses to standard diagnostic testing, thereby improving rates of EGC testing, David Hoffman, MD, a gastroenterology oncologist and medical director of Cedars-Sinai Cancer Beverly Hills, said in an interview. “I think that there are ethical questions that we’re going to have to grapple with respect to accessibility and data mining and what that really means,” said Dr. Hoffman, who wasn’t involved in the study. “But I think that in the optimistic view, using AI with machine learning and deep learning has tremendous potential for public health and for cancer medicine in particular.”
Dr. David Hoffman
Dr. Hoffman added that beyond early detection of cancer, AI systems may hold additional benefits, particularly in regard to assisting decisions for personalized medicine and assisting in real-time surgical interventions. “Using AI with machine learning and deep learning has tremendous potential and I think it sort of is a natural offshoot into what we’re seeing in ... algorithmic approaches to use of big data, so it’s sort of a natural evolution in terms of medical applications,” he explained.
Anuj Patel, MD, a medical oncology specialist at the Dana-Farber Cancer Institute, Boston, who wasn’t involved in the new research, explained that any strategy that helps detect gastric cancer at earlier stages could have a visible global impact. “We have a lower incidence of gastric cancer in the United States, but the training of and clinical volume of early gastric cancers seen by different providers can vary,” Dr. Patel said in an interview. “AI systems could provide a second layer of evaluation during procedures where an endoscopist might otherwise miss the subtle features associated with some early gastric cancers.”
While the findings from the ENDOANGEL study appear promising, Dr. Patel noted that the most important long-term question is whether this reduction of endoscopic blind spots as a result of implementing the system will translate to meaningful improvements in patient outcomes. “More broadly, I think that we will need to see how well these techniques can be applied across different populations,” he added. “Because AI models such as these need to be trained, it will be important to see if they need to be retrained when used in countries where gastric cancer might present differently or with distinctive endoscopic equipment and techniques.”
The study authors, as well as Dr. Patel and Dr. Hoffman, had no conflicts of interest to disclose.