Study in Poland: AI degrades the skills of doctors in tumor detection

With the introduction of artificial intelligence (AI), doctors seem to become less efficient to detect tumors in the colonist themselves. This is what a study carried out in Poland published on Wednesday, one of the first to mention the risk that this technology harms the skills of specialists. The regular practice of AI seems to have “harmful effects on the skills of endoscopy specialists”, summarizes this study published in the “Lancet Gastroenterology & Hepatology”.

More broadly, this is one of the first studies to seek to answer a crucial question: what effect can have AI -based tools, more and more used in the world of health, on medical practice? To answer them, the authors studied data from several Polish centers specialized in endoscopies and colonoscopies, examinations which make it possible in particular to detect digestive cancer signs, in particular the colonist.

The data was collected in 2021 and 2022. During this period, these centers generalized the use of AI software which aims to help specialists to better detect this type of tumors. The researchers did not examine the results of the exams carried out using the AI. Rather, they looked at what those who continued to be carried out by the same specialists gave, but without assistance.

Before the introduction of AI, 28.4% of these examinations led to the detection of an adenoma, a benign tumor but which can potentially degenerate into cancer. Once the AI is generalized, this rate dropped to 22.4%. This therefore suggests, according to the authors, that the use of AI has degraded the capacities of specialists to identify the tumors concerned.

However, the study does not allow you to be sure: it is possible that, over the same period, other factors that AI has played on the rate of tumors detected. “The fact remains that endoscopy specialists would be wrong to neglect the results of this study,” warns, in the same issue of the “Lancet Gastroenterology & Hepatology”, the specialist Omer Ahmad who did not participate in this work.

For him, this study is a first alert, which certainly requires confirmation, on the dangers of AI in matters of “slow erosion of fundamental skills”. “These results nuance the current enthusiasm to quickly adopt technologies based on AI,” he concluded, stressing that this is the first real life study which goes in the direction of a loss of medical skills.