It’s all in the research: Using AI to solve issues in health care

The University of Alberta uses SAS Viya to help its researchers expand their capacity for big data analysis and support the use of open source software and other tools popular among students.

Conducting research is not a straightforward process, and the terabytes of data cascading into labs (both physical and virtual) requires serious horsepower to analyze. Personal desktops and small servers are increasingly coming up short in meeting the demands of artificial intelligence and machine learning projects.

Data also comes in various shapes and sizes. Researchers often combine data related to diagnostic imaging, risk prediction, clinical trials and much more. The need to bring them all together under a single platform for analysis is top-of-mind across the research community. But so is platform affordability for the organization’s finance leaders, particularly when it has tight budgets.

Amid a global pandemic, the University of Alberta quietly launched a new health data management and analysis platform called the Data Analytics Research Core (DARC). DARC increases research capacity and provides high-performance computing and data storage in a secure environment.

Dr. Lawrence Richer, project lead and Vice Dean of Research (Clinical) and Alberta Health Services Chair in Health Informatics Research, said the concept of a shared research platform came to him years ago.

“Open source wasn’t a free option, it was another option,” Richer said. “And when we looked at the standards that we would value in terms of training our students, we looked at the reputation that SAS has for meeting security and privacy standards. For the purposes we were trying to meet, we needed a supported platform. In our province, SAS is very much a standard in the health space. And there were no closed doors with SAS Viya. People who were pursuing other programming languages were not blocked out and could still benefit from a supported platform that met our needs.”

DARC has also already helped accelerate research centered on children with sudden neurologic symptoms. Researchers are working to develop an algorithm that would help reduce CT scans by at least 30%. While CT scans are an effective way to diagnose sudden neurological symptoms, a single scan is the equivalent of approximately 200 chest x-rays. Most people are exposed to this amount of radiation through natural sources over seven years.

“For a young child, that’s a lot,” Richer stated. “I used the automated machine learning tools in SAS Viya, which helped fine-tune and choose the best model. Previously, I’ve also had a machine learning analyst hard code the analysis in Python. What I found was that the automated tuner in SAS Viya did just as well.”

Bringing in the right partner

The University of Alberta was seeking a scalable multiuser environment that had the necessary processing power to accommodate huge amounts of data. It also needed a mix of statistical, visual analytics and predictive modeling capabilities.

Pinnacle Solutions was quick to respond to the university’s needs. Well-versed in business intelligence, data management and predictive analytics – and highly engaged with health care organizations – the SAS partner helped the university design a platform that was anchored to a widely used programming language in Alberta’s health system and played nicely with other open source research software such as Python.

“SAS Viya is a good way to converge business analysts and students coming out of university,” explains DJ Penix, President and CEO of Pinnacle. “Open source has its benefits for rapid prototyping … but you can also proof things out and bring it all together with the SAS Viya platform. SAS, which invests a lot in R&D in its analytic tools, is thoroughly tested and meets regulatory requirements.”

Approximately 40 researchers from the university are registered to the on-premises platform today. DARC also features machine learning tools including TensorFlow, Keras, Caffe and CUDA, which are provided in the Lambda Labs GPU server.

Penix is confident that the university is set up for long-term success. A lot of that has to do with the platform’s ability to enable people across the university’s research teams, regardless of their role, background or job level, to be empowered and get actionable insights from their data.

“As your data grows, your solution should too,” Penix says. “DARC brings together a variety of end users through a common interface, which is something that hasn’t been available before. We're just scratching the surface.”

More results

With DARC up and running, Alberta Health Services was recently able to transfer terabytes of diagnostic imaging data to the platform for analysis.

“We’ve never been able to do that before,” Richer says, describing the entire project as a “non-starter” prior to DARC’s inception.         

Another research project led by Dr. Daniel Baumgart, professor and Director of the university’s Division of Gastroenterology, aims to use DARC to deliver personalized therapies to Canadians living with inflammatory bowel disease (IBD). Researchers are currently analyzing health data from 60,000 Albertans who live with the disease to identify long-term patterns.

Building an analytics platform? Here are things to consider.

The benefits of a modern research platform go beyond the obvious horsepower boost available to researchers and the promise of a secure digital environment to analyze health care data. Some universities that create algorithms can license them to health care providers and other institutions. And tapping into research platforms like DARC can help students get placed into better jobs.

Penix has some advice for other universities and institutions looking to develop an analytics platform.

“First, make sure you understand how vendors define their analytic capabilities,” he says. “In addition to the maturity and sophistication of the analytic algorithms, many niche providers and open-source platforms offer only a small subset of tools to support the entire analytical life cycle. For example, good analytic platforms also include data management and data quality as part of their solution. Second, make sure the platform is scalable. Third, make sure the platform gives people throughout the organization access to analytics, regardless of their role, background or job level. This empowers the entire organization to get actionable insights from their data.”

Richer concurred, adding students’ needs must also be a top priority.

“My advice to other universities considering a research platform is to consider the types of tools that their students will be using in the real world,” Richer says. “Also, the flexibility of tools, like SAS Viya, to meet the needs of people with varied skill levels is crucial.”