Carried out in Sweden and published in the Lancet Oncology, this study above all makes it possible to conclude that there is no risk for radiologists to use artificial intelligence (AI) software to better direct their analyses.
The researchers divided some 80,000 women into two groups of similar size. All performed a mammogram but the first group was screened in the classic way, that is to say with the gaze of two independent radiologists, while the data of the second were first examined by an AI and then by a single radiologist. In the end, the AI-assisted group did not perform worse: they even detected slightly more cancers. The rate of “false positives”, that is to say cases where the first examination wrongly suspects cancer, was similar. Since only one radiologist is needed in the procedure involving an AI, the use of this technology could eventually reduce the workload of these doctors by half.
Promising results
These results are promising because screening is widely considered to be one of the main ways to fight breast cancer. In France, it is widespread among women aged 50 to 74, in line with European recommendations.
However, it is too early to conclude on the real interest of AI in the field: it will take several years of hindsight to know if it has been as effective as a human double opinion. To do this, the researchers will compare in two years the rate of cancers that will have escaped screening but will have been diagnosed in the interval. These initial data also leave room for uncertainty about the risk of “overdiagnosis”, namely of identifying lesions that would not have evolved into dangerous cancers without treatment.
Overdiagnosis?
This question of overdiagnosis is at the heart of certain criticisms of the merits of generalized screening, even if research confirms more and more clearly its interest in reducing mortality from breast cancer. The risk of overdiagnosis “must encourage caution as for the interpretation of the results”, warned in a comment to the Lancet Oncology the oncologist Nereo Segnan, a stranger to the study, while acknowledging its promising nature.