From pre-surgery brain scans to genomic-enhanced diagnosis, high-performance computing is beginning to transform the practice of medicine.

At leading-edge facilities, HPC-enabled advances are already enabling neurosurgeons to get a clearer picture—literally and figuratively—of a patient’s brain tumor. Children born with rare conditions are bypassing years of costly and invasive diagnostic procedures, clearing the way for innovative treatments. Scientists are identifying genetic changes that affect tumor growth, giving cancer patients more effective treatments with fewer side effects.

Precision medicine is not yet a practical, everyday reality—but as these examples show, we’re getting there.

They were among the highlights shared in a recent panel discussion titled Accelerating Precision Medicine with Intel Xeon and Intel Xeon Phi ProcessorsMichael J. McManus, senior health and life sciences solution architect at Intel, chaired the panel. Dr. McManus, who earned his Ph.D. in synthetic organic chemistry from MIT, has worked with everyone from bioscience and technology entrepreneurs to Fortune 500 giants over his 30-year life sciences career. For the panel, he brought together clinical and technology leaders to discuss the progress being made to deliver clinical value through precision medicine—and the further work that is needed.

Speeding time to life

Product developers talk about time to market. Web service providers measure time to first byte. For James Lowey, the key metric is time to life. Lowey is CIO at the Translational Genomics Research Institute (TGen), a nonprofit focused on turning genomics insights into faster diagnostics and treatments that are more effective. TGen’s genetics research is being applied to rare childhood diseases, cancer, neurological disorders, diabetes and others.

“We’ve got patients waiting,” Lowey told the panel audience. “We need to diagnose and treat them. They need results now, not in weeks or months. We’re working to accelerate the movement of insights from the bench to the bedside.”

It’s no surprise that each new generation of processors helps organizations like TGen deliver genetic results—and clinical answers—more quickly. Lowey described TGen’s farm of Intel Xeon processor E5 v3 based Dell blade servers based on Intel Scalable System Framework (Intel SSF). Using the blade servers, TGen has reduced processing time for critical genomics processing tasks from two weeks to seven hours, making it fast enough to be clinically relevant. 


TGen CIO James Lowey discusses HPC’s role in turning genomics insights into faster diagnostics with Intel’s Michael Mcmanus (left), BGI’s Director of the Bioinformatics Center Fang Lin (center), and Boston Children’s Hospital director of the Computational Radiology Lab Simon K. Warfield (right).

A smarter view of the brain

Digital imaging is another area where HPC-enabled speedups are advancing clinical care. Panelist Simon K. Warfield described innovative imaging techniques his team is applying to increase understanding of the brain’s complex circuitry. Dr. Warfield is the Thorne Griscom Professor of Radiology at Harvard Medical School and the founder and director of the Computational Radiology Lab (CRL) at Boston Children's Hospital.

CRL is an Intel Parallel Computing Center that is modernizing the algorithms and data structures of medical image computing on Intel Xeon and Intel Xeon Phi processors. The lab is improving cache performance, vectorization performance and multi-threading performance, as well as creating more sophisticated imaging and modeling strategies. CRL can contribute to improved diagnosis and treatment of brain injuries, multiple sclerosis, depression, Alzheimer’s and many other conditions.

Consider the novel technique CRL has developed to show more clearly water’s diffusion through the brain—and pinpoint hindrances and restrictions to its flow. In contrast to traditional image processing approaches, CRL’s diffusion-weighted imaging infers new parametric maps from data measurements. Its computational model includes tens or hundreds of 3D images—each up to 10 million pixels each—as its inputs. “This type of analysis is very computationally intensive,” Warfield said. “With the accelerated algorithm and the Intel Xeon Phi processors, we reduced the time needed from 48 hours to 15 minutes of calculations.”

That speedup can translate to immediate benefits in for critically ill patients facing brain surgery. That’s because, as Warfield put it, “When you’re talking about surgical planning, life is a matter of time.”

Recently, one of the hospital’s neurosurgery teams realized on a Friday that their patient’s conventional magnetic resonance scan was not clear enough to allow them to proceed with a planned brain resection. With the surgery-planning meeting scheduled for Monday, they requested emergency use of CRL’s diffusion imaging algorithm. The patient had a new scan Saturday evening, the data was processed on Sunday, and the information was ready for the team’s decision on Monday.

Global leaders

The panel also highlighted precision medicine’s global reach—and its big data challenges. Fang Lin, Director of the Bioinformatics Center at BGI, described BGI’s use of the Lustre file system to help maintain storage performance as its data volumes grow. BGI is a global research leader as well as a provider of genetic testing products. It also operates the China National Genebank, putting it on the forefront of China’s five-year. BGI cranks 20 terabytes of sequencing data every day. The institute stores13petabytes of genomic data and uses a 10 petabyte file system comprising Intel Enterprise Edition for Lustre Software and open source technologies.

Dr. David Torrents, a molecular biologist and research professor at the Barcelona Supercomputing Center, shone a spotlight on the importance of collaboration in advancing precision medicine. BSC provides resources to a variety of international centers and consortia. In addition, the institute conducts its own multidisciplinary research in computational biomedicine and related fields.

BSC’s alliances also encompass a range of hospitals and medical centers, enabling it to validate and test its models and tools with data from clinical institutions. “We’re at an exciting moment,” Torrents said. “We are not just developing new solutions for personalized medicine, but now are beginning a pilot program in January 2017 to bring them together and apply them in clinical settings, beginning in Catalonia and then throughout Spain.”

Bridging data silos

The panelists say continued leaps forward in precision medicine will come from faster and more sophisticated analysis of larger volumes of more varied data types. 

“What we want is a more holistic picture, and for that, it’s becoming absolutely critical to combine many diverse data types together for analysis,” said Lowey. To achieve that holistic picture, researchers want to use deep learning and other forms of artificial intelligence. They also want to apply those AI methods to genomic data in combination with imaging data, lifelong clinical records, population studies, environmental studies, and much more.

Different aspects of the precision medicine workflow will have varying processing and storage requirements. So the push continues for faster performance with agile or heterogeneous platform architectures rather than a single “silver bullet” approach. The processors will continue as the primary workhorses, supplemented by embedded resources and FPGA accelerators for parts of the workflow. Distributed compute and storage resources will remain crucial, along with advances in applications and tools.

As to the clinical impact of these holistic approaches, look no further than Boston Children’s Hospital. Noninvasive prenatal genomic testing can indicate whether a fetus has the risk factors that predispose it to be born with a malformed heart. If genetic testing shows these factors are present, data-intensive digital imaging can reveal whether the heart is actually deformed. By combining genomic with other medical data in this way, clinicians can provide peace of mind for worried parents-to-be, or help them plan for their child’s future.

“We’re starting to connect the genetics that predisposes an individual to heart disease, with the imaging to see if the defect is present, and use that information to influence current treatment,” said Warfield. “That information can also help us plan for the child’s longer-term future. We can predict how they’ll do as teenagers and begin to plan accordingly.”

Predictive medicine on the horizon

Precision medicine is one of the most promising and meaningful applications of high-performance computing today.

“It’s still early days, but we’re moving toward an exciting new era of predictive biology and personalized medicine,” said McManus. “Our panelists gave us a great taste of what’s on the horizon. With continued advances in platform technologies, artificial intelligence, and other areas, we create significant opportunities to increase the science of medicine and ultimately improve human health. Intel is excited to empower scientists and clinicians with technology innovations, resources and expertise as we collaborate to make this new era a reality.”

Want to dig deeper?


Jan Rowell writes about technology trends in HPC, healthcare, life sciences, and other industries.