Cancer diagnosis: the France does not want to miss the boat of AI

Among the little or unknown medical specialties: anatomical pathology. Anapaths play a key role in analyzing organs, tissues or cells to identify and analyze disease-related abnormalities. Their mission is particularly crucial to determine the diagnosis of cancer and the treatments to be considered.

These doctors could soon count on the valuable reinforcement of artificial intelligence (AI).

While most pathologists continue to use the microscope exclusively to study glass slides from biopsies, the use of digital technology is increasing.

However, once digitized, this data constitutes a raw material that can be interpreted via AI algorithms and mathematical models.

"The slides contain a lot of information about cells, not all of which are decipherable by the human eye," says Fabrice André, director of research at the Gustave-Roussy Institute, an anti-cancer center south of Paris.

"By reading an image, a doctor cannot predict whether a cell will be more or less sensitive to a treatment, whether there is a risk of mutation ... Whereas AI, yes," he continues.

Four years ago, a study conducted by Gustave-Roussy and Owkin, a French-American start-up specializing in artificial intelligence applied to drugs, showed that AI could designate, among patients with localized breast cancer, the women most at risk of metastatic relapse within five years.

"By reading an image, a doctor cannot predict whether a cell will be more or less sensitive to a treatment, whether there is a risk of mutation ... While AI, yes," says Fabrice André, director of research at the Gustave-Roussy Institute, an anti-cancer center south of Paris © Andreas SOLARO / AFP / Archives

With these results, the anti-cancer center and the start-up become unicorn thanks to the support of the pharmaceutical giant Sanofi, launch Friday a consortium ("PortrAIt"), which aims to make the France "a world leader" in the use of AI to diagnose and treat cancers.

"Never replaced"

The consortium, which includes the Léon Bérard anti-cancer center (in Lyon), the federation of anti-cancer centers Unicancer, the French digital imaging company Tribun Health and the private digital pathology group Cypath, intends to deploy 15 new AI tools within five years. It has a budget of 33 million and is supported by the State via the Public Investment Bank (BPI).

The project actors hope that these tools will help diagnose cancers but will also be able to predict the risk of relapse, which will make it possible to adapt treatments.

"We have already developed two products for breast cancer and colorectal cancer but with this consortium, the idea is to scale up," Meriem Sefta, head of diagnostics at Owkin, told AFP.

On the side of partner doctors, this ongoing revolution is not perceived as competition but as complementarity. "We will always need a +anapath+ to sign the final report of a diagnosis and decide on a protocol with the oncologist, we will never be replaced by a machine," says Philippe Chalabreysse, the CEO of Cypath.

In any case, a handful of players are ready to take on the global market, while the evidence of effectiveness begins to multiply.

Last December, Institut Curie published in a Nature journal the results of a study showing how the AI of the Israeli start-up Ibex Medical had made it possible to detect several categories of breast cancer and determine the degree of severity of tumors.

However, much research remains to be done to prove that the use of AI deserves to be generalized as a tool for detecting cancers. For example, is it effective in people of any age? Does it escape the risk of overdiagnosis?

"It will be necessary to ensure that the tests benefit everyone, to reduce inequalities in care, and that this model is economically viable," warns Philippe Chalabreysse.

© 2023 AFP