AI & the new frontiers of medical radiology, a quick talk with Christian Allouche, co-founder at…

Original article was published by XAnge on Artificial Intelligence on Medium


AI & the new frontiers of medical radiology, a quick talk with Christian Allouche, co-founder at Gleamer

By : Christian Allouche, co-founder at Gleamer

At the forefront of the AI revolution in medical radiology, Gleamer just closed a €7,5 million Series A that will give them all the firepower to address new markets, including the USA. Christian Allouche, co-founder, is up to the challenge.

You are preparing to address the US market after launching Gleamer in Europe this year. What are your next steps?

We will recruit teams, but most importantly we are in the process of launching trials in the USA, which will take a few months to complete. BoneView, our AI assistant for bone trauma detection on x-rays, is a medical device and as such needs FDA approval for commercialization.

We ran similar trials last year in France, which helped secure our CE marking. The study proved that combining our AI with a physician’s eye reduces the rate of undetected fractures by 30%, all while significantly minimizing the time spent reading x-rays. The challenge this time is to prove the efficiency of our solution for American physicians, using local data and on a wider range of x-rays.

How mature is the market compared to Europe?

The american startup VIZ.AI, which analyses neuro MRI scans to help doctors detect and monitor the risk of strokes, just confirmed that Medicare will reimburse the use of their tech in hospitals. This is unheard of, and hugely encouraging. It’s a sign that mentalities are shifting. Healthcare authorities understand that software can benefit patients the same way a molecule would, and that they should back it for this reason.

Our belief is that AI will spread, encouraged by the market, by insurers and healthcare authorities along. Gleamer helps doctors take care of their patients faster and with more efficiency. Its usage is beneficial to everyone involved.

Are there any technical requirements regarding an AI in the medical field?

The challenge is to build an AI that is both excellent and reliable (i.e. which detects as close as possible to 100% of bone fractures). We owe it to doctors, to the healthcare community and obviously to the patients. Simply put, good enough isn’t enough when it comes to medical devices.

There are plenty off-the-shelf algorithms out there. Some even offer interesting results in image recognition… but we had to go further to reach excellence and get an edge over competition. With AI, leaping from good to excellent demands an insane amount of work. We started with existing algorithms that we twisted and adapted to our challenges. It took hundreds of iterations & modifications of our models to build an algorithm that offers satisfying results. Our AI engineers are fighting strong day after day to find fractions of additional performance. It’s a never-ending process, and it’s fascinating!

Read also :