Medical imaging is a broad term that encompasses several distinct technologies. After working on AI-powered tools to enhance X-rays and mammographies, French startup Gleamer now aims to tackle magnetic resonance imaging (MRI).
Instead of starting from scratch, Gleamer has acquired two startups that have already been working on AI-powered MRI analysis: Pixyl and Caerus Medical.
Gleamer is part of the second wave of startups trying to improve medical imaging using artificial intelligence. Several tech founders created startups around this topic in 2014 or 2015. While most of them went nowhere, there have been some consolidation in the space. For instance, Zebra Medical Vision and Arterys were both acquired by Nanox and Tempus, respectively.
Founded in 2017, Gleamer has been building an AI assistant for radiologists, a sort of copilot for medical imaging. With Gleamer, radiologists can theoretically improve the diagnostic accuracy when interpreting medical images.
The startup has already persuaded 2,000 institutions across 45 countries to use its software solution. Overall, Gleamer has processed 35 million examinations. The company has received CE and FDA certifications for its bone trauma interpretation product. In Europe, it also offers products specifically focused on chest X-rays, orthopedic and bone age measurements with CE certification.
“Unfortunately, the one-size-fits-all approach to radiology doesn’t work,” Gleamer co-founder and CEO Christian Allouche told TechCrunch. “It’s very complicated to have a large model that covers all medical imaging and delivers the level of performance expected by doctors.”
That’s why the company created small internal teams focused on mammographies and CT scans. “Three weeks ago we released our mammography product, which we have been working on for 18 months,” Allouche said. It’s based on a proprietary AI model that has been trained on 1.5 million mammographies.
“We have a partnership with Jean Zay, the French government’s GPU cluster,” Allouche said. The company is also working on CT scans for cancers.
But what about MRI? “MRI is a different technological space,” Allouche said. “You have a lot of tasks in MRI. It’s not just detection, you’ve got segmentation, you’ve got detection, you’ve got characterization, classification, multi-sequence imaging.”
That’s why Gleamer is buying two small startups that have been working on this space for several years to move faster. Gleamer isn’t disclosing the terms of the deals.
“These two companies will become our two MRI platforms, with the clear ambition of covering all use cases over the next two to three years,” Allouche said.
Preventive medical imaging
While Gleamer’s models show promising results, they are not yet perfect. For example, with the company’s new mammography model, the startup claims it can detect four out of five cancers. In comparison, a human radiologist without AI assistance typically identifies cancer in three out of five cases.
However, the productivity gains from a tool like Gleamer could radically change medical imaging. A missed tumor is likely to appear in a follow-up exam a few months later.
“In the not-too-distant future, I think we’ll all be getting routine whole body MRIs paid for by our insurance companies — since they’re not irradiating,” Allouche said.
However, in some cities, there are already too few radiologists to meet the demand for reactive imaging. If the industry shifts toward preventive imaging, AI tools will become indispensable.
Gleamer’s CEO thinks AI could become an “orchestrating and triaging” tool. Most medical imaging examinations are conducted as a way to rule out some diagnoses. “So, there’s a real need to automate all this with a very solid AI model that has a much higher level of sensitivity than a human,” Allouche said.
