The use of artificial intelligence in endoscopy can contribute to the non-biopsy diagnosis of precancerous lesions of the gastrointestinal tract so that the correct treatment can be applied and unnecessary interventions can be avoided.

The above mentioned by the gastroenterologist, professor of the School of Medicine, EKPA, Konstantinos Triantafillouin the context of the 43rd Panhellenic conference of Gastroenterology, the work of which concludes today in Thessaloniki.

“The main benefit expected from the application of artificial intelligence in daily clinical practice is to standardize the detection and visual diagnosis, i.e. without biopsy, of precancerous lesions during endoscopy, so that the correct treatment is applied and unnecessary interventions are avoided. Regarding the detection of precancerous conditions, the primary goal of artificial intelligence is to raise the level of performance of less experienced endoscopists,” explained Mr. Triantafillou, and added that “at the moment there should not be too high expectations from visual diagnosis without biopsy to make decisions to refer patients to endoscopic or surgical intervention with the aim of avoiding misuse of surgery”.

According to Mr. Triantafillou, in medicine no test has 100% diagnostic accuracy. In particular, in diagnostic tests where the human factor is decisive for the result, artificial intelligence is applied with the aim of reducing human errors.

“Despite technological advances and the very good quality of training of gastroenterologists, endoscopy of the digestive tract (esophagus, stomach, small and large intestine) is not infallible. It has been observed, for example, that even the best endoscopist with excellent equipment may fail to detect up to 25% of polyps in the colon. Of course, in the majority of them, these polyps are innocent, but some could develop into cancer. Even more worrying – but rarer – is that even cancer is not diagnosed during colonoscopy. Similar alarming findings have been reported for damage to the stomach and duodenum,” stressed Mr. Triantafillou.

As he said, artificial intelligence helps the endoscopist, in real time during the endoscopy and with the help of appropriate software, to detect and characterize lesions of the gastrointestinal tract with or without the prospect of developing into cancer.

“It is important to clarify that the position of artificial intelligence is not to replace human intelligence, but to assist it. Recent studies have shown that when endoscopist skills and machine characteristics are not taken into account, the results of applying artificial intelligence are disappointing,” concluded Mr. Triantafillou.