An Online Wellness Magazine produced by The University of Texas Health Science Center at Houston (UTHealth)

Precision Health

Solving health care mysteries with artificial intelligence

Computer scientists such as Luca Giancardo, Ph.D., of the Center for Precision Health, a joint venture between The University of Texas Health Science Center at Houston (UTHealth) School of Biomedical Informatics and UTHealth School of Public Health, are using artificial intelligence, or AI, to solve medical mysteries.

One of those mysteries is how to detect the warning signs of Parkinson’s disease, an age-related neurodegenerative brain disorder that is expected to rise sharply as the population ages. The gradual damage often goes unnoticed until signs such as rigidity or resting tremors occur. With early intervention, doctors may be able to slow the symptoms.

“It’s not practical to test everyone for the onset of Parkinson’s disease even if you could,” Giancardo says. “It would be better to monitor an activity that people already perform on a regular basis for clues.”

With that in mind, Giancardo used machine learning techniques (a branch of AI) to analyze the typing signatures of people. “The measurements include the time from when you press a key until the time you release it,” he says.

Typing signature

It turns out that people with Parkinson’s have a different typing signature, which Giancardo believes could one day aid in the diagnosis of this condition affecting as many as 1 million people in the United States.

While still in the research phase, the most immediate application might be to see how well people with Parkinson’s disease respond to medications, he says. “Changes in a patient’s typing signature could indicate that a medication isn’t working,” he adds.

Giancardo focused on Parkinson’s disease because its motor signs are relatively well understood and the test is most likely to have a large impact.

None of this would be possible without the high-powered computers that allow Giancardo and his colleagues to detect trends and patterns. “Machine leaning is a specialized branch of AI in the sense that it allows a computer to learn how to perform a specific task. Those tasks include playing chess or finding a kitten in the image,” he says.

Giancardo was recently recruited by the director of the Center for Precision Health, Zhongming Zhao, Ph.D., whose charge to all faculty members is to focus on “actionable” medical issues using biomedical informatics approaches. Zhao is a professor at UTHealth School of Biomedical Informatics, and he has a joint appointment with UTHealth School of Public Health.

“Personalized approaches to manage and treat disease are critical to advancing health care delivery.  An important component to such approaches is our ability to work with large data sets to understand disease and craft strategies for targeted treatments,” says Michael Blackburn, Ph.D., executive vice president and chief academic officer of UTHealth.

UTHealth’s Center for Precision Health is tackling this issue head on by bringing researchers and physicians from all over UTHealth to work as a team to advance approaches toward personalized medicine. From population-based genomics to smart clinical trials, their efforts will help to keep UTHealth at the forefront of this technology, adds Blackburn, the William S. Kilroy, Sr. Distinguished University Chair in Pulmonary Disease and the John P. McGovern Graduate School of Biomedical Sciences Endowed Distinguished Professor at UTHealth.

Blackburn also is a dean of the MD Anderson Cancer Center UTHealth Graduate School of Biomedical Sciences.

At an individual level

“There is a massive amount of biomedical and health care data being generated every day and you can’t analyze it all. We instead focus on areas where we believe we can make a difference,” Zhao says.

Instead of having decisions based on the disease globally, or “one-size-fits-all,” Zhao adds, precision health empowers doctors to conduct patient care at the individual level. For example, doctors use genetic signatures to aid patient diagnosis and develop individualized treatment strategies.

Not limited to everyday tasks such as typing, Zhao’s team also analyzes genetic data, medical images and electronic health records.

For example, Zhao’s research focuses on “actionable” cancer-causing mutations. Many millions of genetic mutations have been identified in cancer. However, not all mutations cause cancer and treatments are available for only a few of those causing cancer.

“We’ve recently developed a powerful approach for identifying actionable cancer-causing mutations,” Zhao says, referring to a new publication that appeared in Cancer Research. “This study, although based on computational and statistical approaches, could help inform many promising mutant proteins for drug development in rapidly evolving precision oncology.”

The Center for Precision Health was established in January 2016. It has four high-priority research areas:

  • Recognizing novel genes and biomarkers for the prevention, diagnosis and treatment of disease,
  • Identifying cancer mutations that respond to available treatments,
  • Fast tracking the translation of medical discoveries into patient treatments, and
  • Designing clinical trials that are more efficient and less costly.

Passive Monitoring

Giancardo was instrumental in the development of the algorithms used to measure typing signatures.

“We call this passive monitoring. This is the type of test that doesn’t require you to do any type of special activity,” Giancardo says. “No trip to the clinic is required.”

Whenever a person types on a smartphone or a desktop computer, a score could be generated to quantify the patterns automatically discovered by the algorithm in the typing signatures. These patterns are very likely to measure slowness of movement (bradykinesia) or loss of movement control (dyskinesia).

The average person takes 100 milliseconds to press and release a key and Giancardo says a new clinical study could be in the offing.

“My vision is to extend these AI techniques to other conditions such as Alzheimer’s and integrate them with imaging and other clinical tools,” Giancardo says.

comments powered by Disqus

This site is intended to provide general information only and is not intended to substitute for or be used as medical advice regarding any individual or treatment for any specific disease or condition. If you have questions regarding your or anyone else’s health, medical care, or the diagnosis or treatment of a specific disease or condition, please consult with your personal health care provider.