The challenge for this improvement came from the OLVG hospital in Amsterdam, as part of the ACE AI Lab program, an initiative of ACE Incubator.
Balázs Borsos is doing his master’s degree in Artificial Intelligence at the VU in Amsterdam and discovered the OLVG challenge during an ACE AI Lab meetup. “The technology is there to be able to do such things,” he says, “so it seemed like a perfect opportunity to go and try it.” He did this through a network of university groups with his team members Laura Alvarez, Diego Manjarres and Julio Lopez. All of them have a strong interest in AI and computer vision and how they can be applied in healthcare.
Computer vision gives computers the ability to process images in a similar way that humans can. A computer is then able to identify objects, analyze and interpret these images and then make decisions or perform an action. Just as part of the human brain processes information it receives from the eyes.
The team’s name, Panops.ai, is rooted in the ancient Greek words for ‘all’ and ‘sight’ and refers to the mythical creature Argus Panoptes. It was, Borsos says, “a really great match. Applying an all-seeing giant with 100 eyes to a company trying to use computer vision for patient diagnosis.”
Detect and determine
The OLVG’s initial mission was clear: to analyze CT scan images to automatically detect kidney stones and determine their parameters in order to personalize treatment. The team had the freedom to determine how they would approach the challenge. They worked with the hospital’s urologists and radiologists to gain insight into their diagnostic processes and how they could be improved. Their initial ideas, says Borsos, focused on “the most state-of-the-art, complex things that we could do.” But the team soon realized they needed to simplify their approach. It was difficult to obtain the large amounts of data needed to train advanced models for machine learning.
Visualising in 3D
The team decided to limit itself and work with readily available data to create a prototype as quickly as possible. The tool was further improved with feedback from medical professionals. The result was a prototype that can automatically characterise kidney stones, as well as visualise them in 3D. The instrument can show variations in the structure and density of the stones. That was not possible with current methods. Being able to measure its size and structure in more detail means doctors can better predict which treatments will be most effective. For example, some larger stones can be treated with ‘shock waves’, so that invasive surgery is not necessary.
Better results, lower costs
Hospitals evaluate new methods and treatments against specific metrics: better physician and patient experiences, lower healthcare costs, and better outcomes. The Panops.ai team met all these points. Their tool automatically creates and records multiple measurements during a single scan. This saves doctors time and provides them with more accurate data. This reduces the need for patients to undergo treatments and the number of failed procedures. This reduces the costs of care and possibly also the length of hospitalization. Traditional methods provide a simple black-and-white image, but the Panops prototype generates a 3D image that contains additional information, aiding diagnosis and treatment decisions.
Collaboration is key
The team presented their project at an ACE AI Lab event in December. Dr. Carolien Toxopeus, radiologist & innovation specialist at OLVG, complimented them on their hard work and described them as “a step ahead of AI companies that don’t have doctors on board.” Borsos agrees that collaboration between innovators and healthcare providers is critical to moving forward: “We work together and leverage each other’s strengths and expertise to create something that is greater than the sum of its parts. We try to maintain a strong focus on our relationships with doctors and medical professionals, because ultimately we want to help them with our innovations.”
New era in healthcare
OLVG and urology unit leader Dr. Ernst van Haarst will now continue the collaboration with Panops.ai after the first pilot period. Borsos is confident that the team will be able to further develop the detection and treatment prediction models. He’s passionate about the potential for AI technologies to unlock a new era in healthcare. “AI is not there to replace human skills, but to augment them. Machines are great at processing massive amounts of data and finding patterns. We can use that to leverage previous treatments, experiences and the knowledge of others. It’s a great addition to human judgment and the ability to make nuanced decisions.”
For innovative startups such as Panops.ai, the ecosystem of our Smart Health Amsterdam initiative offers inspiring opportunities. Not only to connect with health professionals, but also others working on AI and data science in the LSH sector. Borsos says participating in the Intelligent Health and World AI Summit in Amsterdam last year was “a transformative experience” for the team. “We could really go out there and see for ourselves what’s happening at the cutting-edge projects, and network with professionals and great companies. The ecosystem helps startup teams reach their goals. Especially at these conferences, people are really helpful when they see that someone is passionate about something. Even the Chief Technology Officer of a big company who’s seen it all is actually willing to talk to you and to help you out.”
Make an impact
Borsos is optimistic about how challenges – such as those of the OLVG – can continue to stimulate future medical progress. “We strongly believe in making an impact with AI, and improving healthcare is one of the most impactful things we can do today. We hope other people share this vision, whether that be doctors, students, investors or professors. And we hope more and more people will join us on this path.”