PARIS — What if the GAFA quartet (Google, Amazon, Facebook, Apple) also became giants in the healthcare sector? Google first tried to put on the white coat more than a decade ago. After "Google Health," its online medical record project abandoned in 2012, the company made a strong comeback with its subsidiary DeepMind Health, doing what it does best: collecting and processing data. In this case, the data was that of patients in hospitals, particularly in the United Kingdom.
However, things were more difficult than expected with the Royal Free Hospital Trust, which allowed the company to access the medical history of 1.6 million patients. The agreement was terminated when UK authorities found that the data had not been kept anonymous, and had been used in a broader context than originally planned.
But Alphabet, Google's parent company, has several other irons in the fire. The Baseline project, led by its subsidiary Verily, aims to "map" human health by collecting health data from 10,000 volunteers using connected objects. The latest initiative in the Alphabet galaxy, the U.S. startup Cityblock, in which another subsidiary, Sidewalk Labs, has invested, aims to offer medical services and prevention to Medicaid or Medicare members.
The other Web giants are not far behind. Amazon is moving into health insurance, and Facebook AI Research (Fair), after recruiting Yann LeCun, one of the fathers of deep learning, has hired Jérôme Pesenti, a former head of IBM's Watson program. Apple has also confirmed its interest in the health sector with its Apple Watch Series 4, the first consumer device with a sensor and algorithm that can detect a heart attack.
Did IBM's Watson Health start too early? From 2011 and the victory of its artificial intelligence program in the game show Jeopardy, IBM had been targeting the health market, including cancer research. The program, which functions as a diagnostic aid, has been trained by absorbing much of the English-language scientific literature on oncology. According to IBM, and depending on the type of cancer, its treatment recommendations are in line with those made by doctors in at least 8 out of 10 cases, and even 96% of cases for lung cancer.
Doubts have been arising about the reliability of Watson's advice.
However, since last year, doubts have been arising about the reliability of Watson's advice. One of the flagship projects, conducted with the MD Anderson Cancer Center at the University of Texas, has been abandoned. People at IBM say it's because the system is too young.
"There is a real issue around the representativeness of data," explains Silvano Sansoni, sales manager at IBM France. "Watson is used to treat 84,000 patients, mainly American and European, and its recommendations are less applicable to patients from Asia or Africa. Changing the scale to 1 million subjects will require a lot of resources."
Big Blue now claims 230 hospitals worldwide as users, compared to 55 last year. Although it doesn't have a hospital partner in France, it signed an agreement last July with the French company Guerbet, a specialist in radiology contrast products, to conduct research on liver cancer.
Philips is probably the most active company among the major healthcare industrialists, particularly in the medical imaging sector. The firm has just created an AI expertise center in Paris, where it employs some 30 researchers. The objective is to create partnerships with startups and hospitals.
Tech and healthcare are becoming more integrated — Photo: Piron Guillaume
Meanwhile, Babylon Health, a young British company that has raised 50 million pounds since its creation in 2014 and developed a consulting service based on a smartphone application, promises its users "a doctor 24/7." The world of startups using artificial intelligence to innovate in health is booming: CB Insights has identified 481 deals, for a total of $3.6 billion invested over the past five years.
For startups, one of the challenges remains access to data. This is precisely the originality of Owkin, a start-up that specializes in machine learning and oncology research, which counts Google Ventures an an investor. "We collaborate with public research by implementing the principle of federated learning. This allows algorithms to be exchanged so they can be trained without having to externalize the data," says Thomas Clozel, co-founder of Owkin.
On October 4, the young French-U.S. company announced the launch of a collaborative project with Apricity, another startup that specializes in fertility research. The project is called "Substra", funded to the tune of 10 million euros by French public investment arm Bpifrance. The ambition is to develop, together with six major public research institutes, a secure treatment platform and thus to run their predictive algorithms in their preferred fields: anatomopathology, dermatology, and fertility.