Digamma.ai CEO Q&A Series: Wout Brusselaers of Deep 6 AI

Leave a comment
Artificial Intelligence / deep learning / Machine Learning
deep 6 ai digamma ai

Digamma.ai CEO Q&A Series: Interview with Wout Brusselaers, CEO of Deep 6 AI

1. Deep 6 AI recently won the Enterprise and Smart Data category at SXSW’s accelerator competition and participated in the inaugural SXSW Connect to End Cancer event. What role is Deep 6 taking in the fight to end cancer?

We’re not doing any cancer research ourselves. However, Joe Biden, our former Vice President, actually pitched the purpose of our company himself when he was on stage at SXSW. He mentioned that in order for the mutual effort to cure cancer to have a fighting chance, we need to increase the number of patients with cancer that participate in clinical trials. Today, that stands at 4%—only 4% of cancer patients actually enroll in clinical trials. To be able to find a cure for cancer it is widely estimated that that number needs to grow to at least 25% and ideally 50%. Most patients don’t know about clinical trials.

Finding the appropriate clinical trial they are eligible for requires an in-depth, professional understanding of their specific condition, sometimes down to the molecular level. This requires a lot of research, expertise and time. We cannot expect patients to carry this burden to figure out which trial to enroll in, from possibly hundreds available to them, and reach out to all the pharmaceutical companies, the CROs, or the hospitals involved. But trial sponsors or sites also don’t have the tools to look into patient data and quickly assess whether a patient will be a suitable candidate for a given trial. So, that is where artificial intelligence can really help and perform that matching, reducing the time it takes from months to as little as 10 seconds.

After our SXSW presentation, a woman who had Stage 3 breast cancer came up to me and explained how she was looking for a trial to help fight her cancer. Her husband was an oncologist, she was highly educated, and she knew her way around medical data, but she was still overwhelmed trying to find a trial. She said that when she keyed in her basic characteristics she found 140 trials online. She then had to narrow it down by reading through them and reaching out to the different PIs and sponsors. One sent her to a website where she had to take a survey, then a phone interview, and then weeks later they told her that she was not eligible. And she doesn’t even know why. It’s a nightmare. We want to help change that.

2. What is the most transformative impact Deep 6 AI’s technology has had so far?

We’ve been doing this for a little over a year right now so I feel the most is still to come but, for us, it was really the moment that I just mentioned. She made it real. Once you have a patient tell you the struggle that she is going through and how important it is, it really makes you think and reflect on it on a deeper level. It made us realize that we have a tremendous opportunity and obligation to help patients and not disappoint them. So that was a reflection on a personal level for the whole team and, I believe, a very important moment for all of us.

3. Tell me more about Deep 6 AI’s powerful core product, AI-enabled patient-trial matching.

We apply artificial intelligence to all the medical information for patients regardless of the source, including EMRs, pathology reports, laboratory results, tumor registry results, everything. Even devices. We then scan all of that data, structured and unstructured, for the medical content and the most relevant clinical terms. From that, we build a graph that represents that patient.

So, we turn ‘traditional’ document-based, relational database representations of a patient into a rich graph that can display tens of thousands of clinical features and shows all the relationships between those features. The graph can be searched by an algorithm in milliseconds to match against clinical trial eligibility criteria. So, using the power of that AI-enabled patient graph we can find patients for trials in minutes whereas today it takes often months to build a cohort for a trial.

4. How did you identify the need for more effective patient-trial matching? Why did you decide to tackle this problem?

We participated in the Cedars-Sinai Healthcare Accelerator powered by Techstars last year. When we went into the accelerator, we had built a prototype of the patient vector, the dimensional graph. But we decided to use it more for our layman’s interpretation of “AI in healthcare” – clinical decision support. Essentially, we would build a graph for a patient and then compare that against other patients, with very similar graphs, but who are, for example, three or six months further down the line. This would provide us with a ‘crystal ball’ to tell us what this patient would look like in the future compared to other patients very much like them.

We thought it was brilliant (laughs). However, after talking to more than 300 physicians, researchers, and healthcare professionals, we figured out that clinical decision support was a very hard sell for a small company like ours, being a newcomer to healthcare. It would have taken us a lot of testing and a lot more medical expertise than we had on the team, before we could go to market with a product like that.

So we had to reconsider where we could we really make a significant impact with the patient graph that we had built. We wanted to see where we could have a clear impact and ROI for our client based on whether it’s a pharmaceutical company, or CRO, or even an individual patient. After some analyses, we decided that clinical trial recruitment is a neglected but very important problem that needs to be solved. Clinical trials are a critical part of the development and adoption of every new treatment, new product, or new process. By speeding up clinical trials recruitment , we felt that we could help accelerate overall healthcare innovation. I guess we like setting lofty goals for ourselves (laughs).

5. What are some of the biggest challenges you and your team have faced since starting your company?

There are strategic challenges and tactical challenges. Every startup must first figure out product-market fit. The Techstars accelerator really helped us close in on clinical trials as the ‘problem to solve’ for us, and build a prototype, within two months. A larger challenge, which we’re still facing today, is the slow pace in healthcare. Since anything you do in healthcare has a potential to affect many lives, for the better or for the worse, decisions are naturally made carefully, after much deliberation. I.e. adopting innovation occurs at a slower pace than in many other industries. We have to make sure that we are prepared for that, which meant taking a different look at the financing structure of our company and our sales cycle.

6. What advice would you give to companies looking to apply machine learning-based technologies to problems in the healthcare sector?

7. I would advise them to find strong partners, like we did with Cedars-Sinai, to learn from their needs and insights and leverage their feedback as you develop your product. Our access to smart, passionate physicians, researchers and IT people at Cedars-Sinai has been an incredible resource at so many levels. It truly accelerated our understanding of the space and made sure we developed a product that would people would actually use.

7. How do you see AI evolve and transform the healthcare sector in the next decade?

I think it’s going to have a huge effect. However, it’s important to understand that today we’re still talking about narrow AI. We’re not talking about general AI with an all-knowing kind of capacity and self-awareness. In the next decade we’ll mostly see local applications of AI – narrow AI – like NLP and deep learning. At first, these will basically be implemented as point-of-care solutions, informing individual medical decisions like, “Can we move these patients out of the ICU and back to a normal bed?”

But I believe that, over the next decade, we will starting seeing a more holistic approach where AI can be used to analyze a patient’s health across multiple sectors like lifestyle choices, care choice, treatment choice, maybe even professional choices. From there, you can have AI-driven decision support to reconcile the different states of your life and make sure that the overall output somehow optimizes your health.

With over a decade of AI experience, Digamma.ai’s team are your trusted machine learning consultants, partners, and engineers.


Leave a Reply

Your email address will not be published. Required fields are marked *