December 11, 2019
Top Five Podcast: Dan Housman, Founder & CTO of Graticule
In this episode of “Top Five” we catch up with Dan Housman, Founder and CTO of Graticule. Top Five is a podcast series recorded at RSNA.
Q: Dan, thank you so much for joining us today for this podcast. You are the co-founder and CTO of Graticule, explain to our audience what Graticule does?
A: Graticule is a real world data company and we focus on advanced real world data. Those are the sources that have been traditionally very hard to access for life sciences companies, device companies, and AI groups. It includes things like imaging data, genomic data, and free text notes, all the things that are hard to de-identify or process. We’ve been trying to build out because those are the areas where we see there’s the highest value in terms of what’s richly in the content in those data sets as well as places where we can apply some of this advanced computing that recently has come about, like natural language processing, machine learning, and AI.
Q: Everyone’s talking about healthcare data on multiple fronts from security, privacy, patient control, how to advance it, how to deploy real world evidence, in your estimation, what are the top three things that need to be done in order to advance the use of data?
A: I think a lot of the issues are in the maturity of the market. There’s still a huge gap between what a life sciences company or a group that’s going to be consuming real world data want to get done and understand about what’s available and the groups that have that type of information. So, probably the biggest thing that needs to be filled in is the gap of communication so that there’s a much easier dialogue on how to contract data between these two groups, what is acceptable use of what kinds of data can be contracted, what kind of analyses can be done behind the firewall versus on the site within the life sciences company.
I think there’s a there’s a whole lot of work to be done on the business model. I think some of that is still in the hands of changing dynamics on the regulatory front. What is going to be allowable tomorrow may not be exactly what’s allowable today. I think a lot of the winds are shifting more towards increased restrictions in terms of flexibility around privacy, technology’s going to have to come through with some solutions to be able to scale how we can get the science done we need to do without putting any of the individual patients that are involved in harm’s way within the data sets were working with.
I think another big piece, from a more global standpoint, is that things have started to change with regards to how well you can distribute consent. What I mean by “distribute consent” is, it used to be the only way to do informed consent would be you come in with a very specialized visit at a health system, sign a document, and then you would be consenting to commit your information into research. Now we have mobile consent, which means you can use those things we like to look down and stare and all day, our phones, to consent to things. A shift really has to come about if we want to unlock a lot of information. It is going to shift from most information we use in real world data coming from a health system or a clearinghouse to coming from individuals. A lot of the pieces are in place but the see-change-next-step really hasn’t happened yet. I think both cultural changes will affect some of this, maybe some new technology, and maybe just how the business of investing and getting people to consent to these large scale studies really come about.
Q: Once those practical considerations are addressed and you can get your hands on some advanced data that previously were unavailable, what are some of the questions that can be answered? What can we do with things such as imaging data? What kind of questions can we answer that were previously unanswerable?
A: I’ll start with this idea of identifying new biomarkers. The one we’re most familiar with is something like HbA1C. If you took a look at someone’s HbA1C level, which is a lab test, we have a very good idea of how far the progression is of their diabetes. Then we can work towards improving or modulating the A1C level with drugs, diet, and behavior.
Within the images that we have already in the health system, there’s the potential of a whole set of new biomarkers developed when we look inside of things that are already being looked at by standard of care. We have a standard of care for mammography, there’s now standard care for chest CT, and standard of care in a lot of cancer follow-up. With all of these things, we can identify new markers within that dataset and it will really help us to be able to find the right drug for each patient.
One of the examples we’re working towards is, there’s a company with new drugs for multiple sclerosis that treat patients who have secondary progressive. The hope is that we can look inside of historical imaging data for patients who have had MS to see if there are specific changes within the brain. They’re already getting regularly screened and the annual screening that would indicate that those patients will have or do have secondary progressive MS and then we should prescribe the appropriate new drug for it. By putting into place these kinds of new “diagnostics,” whatever we want to think of them as, that comes out of learning from what we already have built out as past information but projecting into new medicines, now we can help treat people in new ways. The next real step in that kind of progression is if we integrate this back into the health system and translate the new biomarkers, new diagnostics into processes like clinical trial recruitment, where it can be really hard to recruit people who have a specific disease, now we can easily identify a group by looking at past imaging records or past deep medical records or genomics. The huge shift is both on building new markers and new diagnostics and then translating them so that we can look at very large amounts of historical data to help patients and help the people who care.
Q: Talk about your partnership with Life Image. Why does that make sense in terms of how we can complement each other to advance the use of more sophisticated data?
A: There’s something we really like at Graticule about Life Image, which is Life Image isn’t just a tool, Life Image is a network. It’s the largest health information exchange for imaging data, and as a result, there are many different nodes that are already collecting information and passing it to each other of imaging data. In the case of working with Life Image, there’s a set of that data that we can use readily for real-world studies because Life Image worked with the providers and got the right contracts in place to allow that to be used for secondary purposes.
We at Graticule are working the end that’s doing the translation between what life sciences companies want, need, or interested in, and what’s really available today or could be made available by diving deeper into the network. On the one hand, what Graticule is doing is navigating within our client base of life sciences companies for very specific use cases. We are looking at if they are interested in echocardiograms, or the new breast cancer diagnostic, or do they want lung cancer to try and see if they can move things from early stage? Whatever the use case is, we’re elaborating it to the point to which we can engage with both the Life Image data we already have as well as with additional information that we’ve brought in from the network. One of the things we’ve done to make this easier is we’ve taken a set of patients that are available for secondary use with their imaging, and created a de-identified snapshot of the information, reducing it down quite a bit using natural language processing and data extraction tools so that we can understand how to build cohorts and do feasibility. We call that GLIMPS, the Graticule Life Image Machine Parse Set. Now we have a portable easy way to analyze a data set that has been able to avoid a lot of the issues we face with how do you de-identify an image fully? Or how do you de-identify text fully? Now we can collaborate with these life science partners to form good questions and be able to structure some of the studies that we want to do, bring it back to Life Image, and then be able to execute these studies and make a big impact on either their product portfolio or the patients that they’re treating.