Google algorithm would be able to screen for breast cancer more effectively than doctors

[ad_1]

An artificial intelligence technique developed by Google has been shown to be more effective than doctors in detecting breast cancer. This is what suggest the researchers of the Mountain View firm in a study published this Wednesday, January 1 in the scientific journal Nature. By analyzing mammograms, the algorithm could overcome the limits of healthcare professionals, who still miss 20% of cases according to the American Cancer Society. No less than one in eight women worldwide will be affected by this condition in their lifetime.

Exceed human capacities

Nearly 30,000 x-rays were used to feed the algorithm in the UK and United States. The study pointed to his ability to identify cancers with a similar, if slightly more, degree of precision than that of six practitioners. What reduce, on average, by more than 5% the number of false positives and 9% that of false negatives. Connie Lehman, head of the breast imaging department at Massachusetts Hospital, noted that "The results are in line with the conclusions of several groups using AI to improve the detection of cancer during mammograms".

Today's computer programs used by healthcare facilities can only locate items that are already detectable by radiologists. The objective of research around AI is to anticipate the occurrence of a disease – such as the technique recently developed by the University of California to simplify the diagnosis of cerebral hemorrhage – or to avoid missing out on it. '' an element invisible to the naked eye on an image. In other words: to exceed human capacities. "The results of the study represent a major step forward, proving that such a tool can really make very informed decisions", welcomed Mozziyar Etemadi, one of the co-authors of the study, which still has limits since the cancers were known in many patients tested.

This finding is the result of research by DeepMind. The company, which is a separate entity in the Alphabet group, recently sold its health division to its sister company Google – which it joined with one of its co-founders, Mustafa Suleyman. If she regularly publishes promising work in artificial intelligence, she also tends to get carried away. In the summer of 2018, she notably touted an algorithm to improve the detection of kidney failure … before independent researchers pointed out the lack of rigor in her research. It would therefore be better to wait for the actual application of these new techniques before gauging their effectiveness.

[ad_2]