News

Home > News & Events > Medial EarlySign Uses Machine Learning Algorithm to Predict Who is at Risk for Certain Conditions

Medial EarlySign Uses Machine Learning Algorithm to Predict Who is at Risk for Certain Conditions

Medial EarlySign Uses Machine Learning Algorithm to Predict Who is at Risk for Certain Conditions
Date

November, 21, 2017

By Erin Dietsche

 

Founded in 2009, Medial EarlySign is taking advantage of artificial intelligence technology and leveraging it to predict which individuals are at risk for specific health conditions. The company utilizes blood test results and EHR data to do so.

 

“Prior to founding the company, we actually spent a few years … investigating what could be done with existing health records and what we can harvest from these kinds of data using different approaches of machine learning,” cofounder and CEO Ori Geva said in a recent phone interview.

 

Most recently, the company used its technique to conduct research on prediabetic patients.

 

In a study based on data of 645,000 prediabetics, Medial EarlySign discovered that by isolating less than 20 percent of the prediabetic population, its algorithm pinpointed 64 percent of individuals who became diabetic within a year. The algorithm ranks the 20 percent of patients based on risk stratification.

 

The EarlySign platform uses over 25 parameters from data stored in EHRs.

 

But the company isn’t limited to prediabetes. Its solution can be used for detection of other conditions as well.

 

For example, EarlySign implemented its technology at Maccabi Healthcare Services, an integrated delivery network in Israel. The tool, ColonFlag, is helping MHS identify patients with a high probability of having colorectal cancer.

 

Oxford University also conducted research on ColonFlag. Another analysis of the tool was based on blood samples and demographic information from Kaiser Permanente Northwest patients. ColonFlag is not cleared by the FDA for use in the United States.

 

“When we work with providers, we can … understand what kind of challenges we might have with these data and how to work with this,” Geva said. “This is really part of our approach — to be able to understand the clinical aspects of what we’re dealing with and also validating this with prominent research partners.”

 

Read the full article here.