How artificial intelligence can detect hidden diseases

As the field of medicine advances, new connections leading to the diagnosis of rare illnesses are being made. Komodo Health’s artificial intelligence algorithms are sifting through a decade of data about health conditions across several hundred million Americans, many scenarios that once were the stuff of imagination are now becoming pathways for the next clinical assessment to be taken.

Komodo Health, founded in 2014, has mapped out 300 million individual health identities across the country to find patterns signaling the presence of disease, years before they are even noticed.

At a time when chronic conditions account for 75 percent of the U.S.’s $3 trillion per year health care spend, identifying symptoms earlier by recognizing patterns of activity that are precursors to those diseases is critical. Not only will this reduce health care spending, it will enhance quality of life and possibly even add years to people’s lives.

In the era of artificial intelligence, predicting the development of disease is now within reach. Not only that, but thanks to genetic testing and personalized medicine, it’s not just the more common diseases getting attention. Rare diseases—often overlooked, disregarded, and orphaned because there’s no profit in creating medicines to treat them—may be getting a second look from pharmaceutical companies.

With the oversaturation of the market for common medications such as treatments for high cholesterol or erectile dysfunction, for example, big pharmaceutical conglomerates are turning their clinical development focus towards the pursuit of treatments for rare diseases.

Can AI help in predicting and creating treatments for rare diseases?

Komodo Health is already making inroads in this area. A particular success was with heredity ATTR amyloidosis, a very rare disease affecting only 50,000 people globally. hATTR is a multi-system disease that affects the heart, liver, kidneys, and nervous system, and once diagnosed, the median life expectancy is five years. Making this disease even more difficult to detect and manage is the fact that there are 120 TTR gene mutations to test for, and very few people with the symptoms of hATTR even get tested in the first place. Because the disease is so rare, physicians often don’t think to test for it until catastrophic illness results from the mutation.

Disconnected symptoms like those in hATTR are often viewed in silos by the medical establishment, but robust AI algorithms can now look across the range of symptoms through multiple years of care and weave together a pattern that indicates hATTR.

Komodo Health used its healthcare map to piece together predictive elements of patients with known hATTR. It then mapped out the clinical profiles of hATTR patients to understand patients’ visits from one care provider to another, across multiple specialties and treatments and diagnoses. From this information, Komodo was able to identify early patterns of hATTR before it progresses to a serious condition.

What the company’s team found was that patients with hATTR had, on average, more than 100 clinical encounters before they were diagnosed. The Komodo Health healthcare map and algorithms were able to hone in on activities and manifestations of the disease at least three years prior to when those patients would have been diagnosed.

Biology and AI advances are opening up new realms of specific and granular data about diseases that have never existed before. In the U.S., rare diseases affect fewer than 200,000 people out of a total population of about 327 million, so finding these individuals is extremely difficult. The only way to find them before diseases affect patients’ quality of life is to have mapped out the journey of others who have had such diseases. Artificial intelligence-based platforms like Komodo Health are validating these new approaches.

Read the full story here.