Working with the Netherlands Cancer Institute, Ellogon.AI has developed a faster and more efficient method of determining whether a cancer patient may benefit from immunotherapy. The tool used for this, EIDOS, is based on big data and artificial intelligence (AI) and runs on the UbiOps platform. As a result, the service was able to go live faster, be offered as an SaaS solution and be easily scaled up as more hospitals – both in the Netherlands and internationally – start using it in the next few years.
This is an important breakthrough in medical science. The more accurate selection that EIDOS enables, ensures more benefits for patients as well, providing a more targeted application of the expensive drugs, explains Robert Kuipers (CEO of the Amsterdam-based start-up). Determining in advance who is most likely to respond well to the treatment is a complex task.
“A pathologist has to interpret tumour tissue with the naked eye and that is very difficult. Rather like looking at the night sky and having to say which stars are between 10 and 40 light years away,” says Robert. To illustrate this, if you show the same tissue to ten pathologists, about three of them will agree.”
Faster and better
“On average, 45 out of 100 cancer patients are selected for immunotherapy. Only 15 of them actually benefit. This is regrettable, because it is an expensive therapy. What is even more important is that people who would have benefitted are currently not selected. We use AI to make that selection process more accurate.”
The Ellogon.AI tool uses a digitised scan of the tissue. “This whole slide image is analysed in a cloud environment by our algorithm, developed with AI and constantly improving. The AI works very precisely — for 24 hours a day, without ever getting tired.”
This not only delivers better results, but also provides greater capacity and saves pathologists a lot of time when examining tissue for the indicative biomarkers. It takes only 20 minutes from the moment when the pathologist sends the digital image of the tissue through the cloud until he receives the biomarker report containing the quantified data from Ellogon.AI. Based on this, the pathologist can give much better advice on whether the patient in question will benefit from immunotherapy.
No proprietary hardware and cloud environment required
Ellogon.AI was founded in 2019 by two professors of Artificial Intelligence at the University of Amsterdam. They recruited Robert in early 2021 to lead and expand the start-up. Fourteen people now work at Ellogon.AI. They are mostly AI engineers and medical scientists.
From the beginning, the company decided not to invest in its own hardware and cloud environment. Nor in an in-house IT department. “The mission of Ellogon.AI is to excel at making and translating medically relevant developments and biomarkers into accurate algorithms. We do not want an extensive in-house infrastructure with massive computing power in the form of Graphic Processor Units (GPUs) to enable workflow and calculations. For that, we use UbiOps’ platform.”
Hospitals also do not need to invest in new hardware and software to use the solution. “Thanks to UbiOps, we can offer our tool in the form of Software as a Service, which works in the cloud. All the hospital has to do is send us the digitised tissue.”
Stable and robust
“Without UbiOps, we would easily have spent six more months on the go-to-market strategy. And then we would still have had to find good full-stack developers if we had had to build everything ourselves.”
“At a company where I was formerly employed, we did it ourselves and so I know all too well how complicated it is and how much money and time it takes to get and keep everything working and stable. UbiOps was able to offer us a platform that was already running in a production environment. It is a robust and very user-friendly environment, into which we were able to integrate our complex workflow well.”
“Thanks to UbiOps, we are also cloud provider-independent. In the event of an outage at a cloud provider, UbiOps ensures that we can switch to another provider straight away and thus continue with our analyses. This means that we will still be able to provide the hospitals that rely on us with their results quickly.”
EIDOS is now CE-certified. “We have been running a pilot environment for a while now and conducted numerous tests. We are now at the point of implementing the solution in clinical practice. Many academic medical centers in the Netherlands, Germany, Belgium and the UK have already shown interest in our solution. The goal is to start with about ten cancer centers in the Netherlands and abroad by 2023.”
It is important that Ellogon.AI can scale up easily. “The volume is definitely going to increase. UbiOps provides a scalable environment. Meanwhile, as an added bonus, we only pay for the computing capacity that we actually use,” mentions Robert.
“What was also important in our choice was the fact that UbiOps is a Dutch company and similar in size to our own organisation. It’s also good to know that I am just one phone call away from the CEO. With a big mega-organisation, you can forget about that. They are able to adapt quickly and are willing to contribute, growing along with us in order to implement our solution. Because their social impact is fundamental to them.”
“Without UbiOps, we would easily have spent six more months on the go-to-market strategy.Robert KuipersCEO at Ellogon AI
And then we would still have had to find good full-stack developers if we had had to build everything ourselves.”