Floy: AI start-up improves evaluation in imaging

Floy from Munich brings AI in imaging into practical application and is already present in numerous German radiology practices. The start-up's software is designed to support radiologists in diagnostics and prevention by reliably detecting critical incidental findings.

Benedikt Schneider, CEO of Floy, explains in an interview with MEDICA-tradefair.com why the solution is very attractive for both practices and patients, what possibilities the additional evaluation of image data opens up for diagnostics and how Floy intends to continue growing.

Mr. Schneider, how does your AI improve the work of radiology practices?

Benedikt Schneider: Radiologists have a clear focus when they perform an MRI or CT scan. The AI points out information that is contained in the image data but is not currently causing any symptoms and is therefore not the focus of the examination.

Around two percent of the population have brain aneurysms, which are blood-filled bulges in blood vessels. These can burst unexpectedly, often without any prior symptoms. As the medical issue is often different from this, the focus of a routine radiological examination is elsewhere. Our AI technology alerts radiologists to changes near the blood vessels.

We are also currently developing screening software for a widespread disease that is unfortunately often underdiagnosed: osteoporosis.

What makes your solution so attractive?

Schneider: The results are very valuable for patients who are diagnosed and treated for serious illnesses at an early stage. Doctors can use AI to improve the quality of their work.

The remuneration model is interesting for both sides. The services of additional evaluation of radiological data are available to self-payers. Practices therefore bear no financial risk and have no additional costs.

Initially, we concentrated on approaching practices and radiologists. However, we are currently seeing an increasing trend of more and more patients requesting our solutions directly. We benefit from the fact that knowledge about the possibilities of AI is also very widespread among patients.

What does the collaboration with the practices look like technically?

Schneider: Image data is exchanged between the practice and our AI via a DICOM node in compliance with data protection regulations. We write the result back to the existing MRI or CT study as an additional series so that it is directly included in the data set.

Which applications are already available and which are planned?

Schneider: Our portfolio includes applications for spinal MRI and CT, abdominal CT, thorax CT and head MRI. Two future topics are MRI of the knee and MRI of the shoulder, because both are very complex joints and vary greatly from person to person.

The development of further applications is all about frequency and the question of where we can add the most value: How well can diseases be detected and how great is the therapeutic benefit?

How good is the data situation for developing and training AI?

Schneider: We are already working with more than 180 practices. MRI and CT are widely used; there are more than 20 million examinations in Germany every year. Acquiring data is not a major challenge.

What is more important is filtering the data that improves our AI. We approximate which new data will give the AI the greatest learning gain. We also ensure that the data basis for devices and personal and disease profiles is broadly diversified.

Where is the potential for growth?

Schneider: Going through the willingness of patients to buy is currently a very good way to go. Billing via health insurance companies is more of a medium to long-term prospect for the future. Our model can be easily internationalized and since we are certified for the entire EU in accordance with the MDR, it makes sense to expand within the EU, especially in the DACH market. However, we are also looking at the American market in particular.