In a new study published in Nature Medicine, Mayo Clinic and several research collaborators from across the U.S., describe a successful new artificial intelligence, or AI, -enabled tool to identify idiopathic pulmonary fibrosis, also called IPF, before patients have recognizable symptoms. This tool could alert a patient's primary care team of a probable IPF diagnosis based on comorbidities and risk scores. The tool should prompt earlier referral to pulmonary specialty care to confirm the diagnosis with CT scanning and lab testing.

In the paper, the team describes how they were able to use "pattern discovery algorithms [to] identify subtle comorbidity incidence, timing and sequence characteristics presaging IPF."

Read more in a news story published by the University of Chicago (@ScienceLife), home to study lead author, Ishanu Chattopadhyay, Ph.D., a machine learning expert whose "research focuses on the core algorithmic principles of large-scale data analysis, particularly in domains where minimal human intervention is warranted or desired, and where domain expertise is scarce."

"IPF is a debilitating and ultimately fatal disease that is often difficult to diagnose," says study co-author Andrew Limper, M.D., a pulmonary and critical care specialist at Mayo Clinic and leading IPF researcher. "Until now, that diagnosis required a barrage of tests. Furthermore, by the time IPF is identified, it's often long after the patient has been struggling with advancing symptoms."

According to the National Institutes of Health, people with this disease usually survive 3 to 5 years after diagnosis.

"We hoped to find an AI solution for earlier identification, because reaching the IPF diagnosis can often take as long as three years after symptoms start," explains Dr. Limper. "Although there is currently no cure for IPF, earlier diagnosis gives patients more options to stall disease progression and maintain an optimal quality of life."

This type of research is common in the Mayo Clinic Robert D. and Patricia E. Kern Center for the Science of Health Care Delivery, where data is an essential element of every project. With expert data analysis, center researchers are able to find answers to questions that arise in the day-to-day of health care, such as:

• Will this medication work well for my 30-year-old pregnant patient? How about her 80-year-old grandfather with diabetes – will he get the same benefits?
• Are there scheduling process changes we can make that will allow us to serve more patients, without overstressing or otherwise harming our staff?  Changes that actually improve the experience for everyone?

Analyzing the care delivery and outcomes information from hundreds and thousands of patients, even millions, can lead to answers that are widely applicable, not just for the participants in a carefully designed clinical trial. By knowing 'all' the answers, to 'all' the questions, a physician could provide an instant, personalized diagnosis and treatment plan unique to each patient — enter the medical tricorder of Star Trek fame.

Although health care technology hasn't reached full Star Trek capabilities yet, this study, and the artificial intelligence and machine learning methodologies it employs, bring the world a step closer.

"Electronic health records, administrative claims databases and ever-growing storehouses of clinical knowledge and research findings provide an endless sea of data from which to derive AI and machine learning solutions to improve outcomes and the experience of health care," says Che Ngufor, Ph.D., a machine learning expert and leading developer of artificial intelligence solutions in the Mayo Clinic Kern Center for the Science of Health Care Delivery.

"We really are working in an amazing time in history — what once was science fiction is our new reality, and often our only limitations are the vividness of our imagination, and the speed and precision of our computers."

Dr. Ngufor works to "develop end-to-end AI and machine learning solutions for descriptive, causality, predictive, and prescriptive analytics with the goal of optimizing clinical decisions and interventions."

And the work that brings a little extra sparkle to Dr. Ngufor's eye? Exploring possibilities with virtual and augmented reality and 'superior artificial intelligence,' essentially aware, thinking, near sentient neurocomputers.

Dr. Limper leads the Lung Defense, Infection and Fibrosis research laboratory at Mayo Clinic. He holds a joint appointment in Biochemistry and Molecular Biology and is also the Walter and Leonore Annenberg Professor of Pulmonary Medicine. His interest in deep data analysis, as well as AI applications for the transformation of health care, stems from his time as the leader of the Mayo Clinic Kern Center for the Science of Health Care Delivery.

Other members of the study team are:

• Dmytro Onishchenko, a data scientist in Dr. Chattopadhyay's Zero Knowledge Discovery — ZeD Lab, at the University of Chicago.
• Robert J. Marlowe, the founder and president of Spencer-Fontayne Corporation, Jersey City, New Jersey.
• Louis Faust, Ph.D., a data scientist who works closely with Dr. Ngufor at Mayo Clinic.
• Gary Hunninghake, M.D., director of Interstitial Lung Disease Program, Brigham and Women’s Hospital, Harvard Medical School, Boston.
• Fernando J. Martinez, M.D., chief of the division of Pulmonary and Critical Care Medicine at Weill Cornell Medicine, New York.

###

Related articles