Bio IT: Getting to the Heart of Traumatic Brain Injury

As Bio-IT World approaches this week, we are sharing a pre-show guest blog series from industry experts on trends you can expect to hear about at the event. Below is a guest contribution from Dr. Pek Lum, Chief Data Scientist and VP of Solutions, Ayasdi.

It seems like every time that I turn on the news, I hear another story about someone with a traumatic brain injury whether they received it on the football field or the battlefield. The fact is that traumatic brain injury (TBI) is something that can affect anyone at anytime. A minor car accident, a trip on the stairs, or a tumble on a bike, can send you to an emergency room. In fact, up to one-third of people hit their head hard enough to go to the hospital by their mid-twenties, according to UCSF. Thankfully, most of those head injuries are mild and patients feel better in a few days. However, that’s not true for roughly 20 percent of patients who go on to develop persistent problems—such as depression, memory issues, or headaches—that can last weeks, months, or sometimes even years.

When concussion patients arrive at an emergency room, doctors diagnose the severity of these injuries based on clinical behavior and CT or MRI scans. But, those scans don’t tell us everything. They can often miss subtle physical injuries and they tell us nothing about which patients will be fine in a few days and which will go on to develop lingering adverse effects.

These are the questions that Dr. Adam Ferguson and Dr. Esther  Yuh at The Brain and Spinal Injury Center (BASIC) at UCSF wanted Ayasdi’s help to answer.

Advanced brain scan data from Diffusion Tensor Imaging (DTI) can reveal brain abnormalities that would not necessarily show up on an MRI or CT scan. That said, the amount of data that is generated from a DTI scan is massive and incredibly complex.  There are millions of questions that could be asked of the data and it takes time and money for Data Scientists to find the right ones, if they do at all. It’s literally like trying to find a needle in an enormous haystack. Ayasdi, on the other hand, is perfectly suited to this kind of task. The Ayasdi Cure application uses Topological Data Analysis (TDA), combined with an ensemble of machine learning techniques, to enable domain experts and Data Scientists to easily discover insights automatically. In fact, Ayasdi Cure is now processing 400 percent faster because the application is now optimized to run Intel® Xeon® processor, including the Intel® Math Kernel Library (MKL) and Intel® Advanced Vector Extensions (AVX), which dramatically reduces the time to insight.

What is abundantly clear is that the labels of mild, moderate, and severe head injury are too simplistic and don’t tell us the full story. By collaborating closely with UCSF and partnering with Intel, we hope to find more sensitive TBI markers that will define more precise patient subpopulations.

By doing this, doctors can potentially predict whether someone will develop complications at the time of the fall or hit.  Knowing this information ahead of time can help both doctors and patients mitigate complicating factors and potentially avoid them altogether.

This vision of the future is not just a dream. We’re making it happen. Our preliminary work is so promising that Ayasdi and UCSF were awarded the GE/NFL “Head Health” award to explore how to better diagnose and treat mild traumatic brain injuries in professional football players.  We are looking forward to sharing our findings in the coming year.

How are you tackling massive amounts of clinical and genetic data?

Dr. Pek Lum is the Chief Data Scientist and VP of Solutions at Ayasdi.

Keep up with Pek @peklum and Ayasdi @ayasdi