Artificial intelligence for relearning to walk to aid in gait rehabilitation after a serious pathology

Leaning on his crutches, Jean-Luc Désire straightens up for a short round trip in front of the "KinTrack" then comes back, curious, in front of the screen: thanks to artificial intelligence, this machine instantly provides a complete analysis of his gait, a valuable new tool for his rehabilitation.

The device was designed in 2019 by the startup Ekinnox, located in a business nursery in the Sophia Antipolis technopole, near Nice.

Baptiste Fosty, a 37-year-old engineer, previously worked at Inria, a public laboratory for computer science and mathematics research, where he developed an image-based measurement tool for the walking speed of Alzheimer's patients. With the blessing of Inria, he chose to develop this technology to aid in gait rehabilitation after a serious pathology: stroke, neurological disease, amputation...

Nicolas Maire, a 36-year-old salesperson committed to a motion analysis project for high-level sports, after a modest career as a professional footballer, joined him to establish Ekinnox, which now has seven employees thanks to initial fundraising of 2.5 million euros.

Their machine, mounted on wheels, relies on a computer analyzing images from a Kinect camera, originally designed to distinguish the movements of Microsoft Xbox 360 game console players. A round trip by the patient in a corridor is enough to provide a 3D projection on the screen, as well as numerical data and graphs on walking speed, step length, foot spacing, weight distribution, hip, knee, and ankle angles.

The Orsac-Montfleuri clinic, a rehabilitation center perched on the heights of Grasse, equipped itself with this tool a year ago, thanks to funding from the departmental council.

Before, gait analysis was "purely qualitative," according to the practitioner's eye, explains Louise Basset, the physiotherapist in charge of rehabilitation. Now, "it allows us to have more precise monitoring," with numerical data quantifying progress that may not be visible to the naked eye.

At 68, Mr. Désire suffers from nerve damage: after five weeks of rehabilitation, the machine confirms that he stands straighter, his steps are longer, a sign that he has regained confidence, the gap between his feet has decreased, a sign of better balance.

Intended for caregivers, the machine has proven valuable for the patient as well: "I can watch the progression of my walk, my posture, and see all the numbers a bit (...). It allows you to see the evolution, it's pleasing," he testifies.

And by showing that his data are identical when he lets go of one of his crutches, the machine encourages him to take the step.

"It's a good monitoring tool," comments Jonathan Bredin, head of the motion analysis laboratory at the Rossetti Institute of Motor Education (IEM) in Nice, where Ekinnox founders came to test their machine and ensure it produced sufficiently reliable data.

Thanks to a mat recording foot pressure and eight infrared cameras reflected on markers attached to the patient—similar to special effects in movies—the laboratory records the movement very precisely and then provides a 3D projection on the screen of the studied limb.

But the entire setup costs 550,000 euros, and it takes a whole day to analyze a single patient, usually for complex surgical decisions.

Less advanced but much faster, the KinTrack, at 45,000 euros, still represents a significant investment. However, Ekinnox has already sold 25 in France and is looking for opportunities abroad.

The startup also hopes to eventually offer a smartphone application if cameras improve. And Mr. Maire has not given up on his idea of developing the technology for high-level athletes.

In the meantime, the small company was surprised to be approached for less conventional applications, such as detecting abnormal behaviors in a crowd or offering a personality analysis based on movement.

"It's not because we can use the software on all topics that we will go there (...). We will always have a vocation for positive movement analysis," explains Mr. Maire.