Figure 1, a social network that allows medical professionals to share and discuss diagnostic images, plans to add computer vision to its service that will enhance the analysis of electrocardiogram data.
The company will introduce an AI-based system that will turn electrocardiogram photos into digital data. The system initially will allow clinicians to give their opinions on the meaning of the measurements. However, in the future, more sophisticated machine learning systems may be able to add a new level of insights into electrocardiogram analysis, as reported by Fast Company.
Electrocardiograms monitor heart activity, making them useful for diagnoses of a variety of cardiac issues. They also can help doctors understand the impact of other health issues on the heart, including high blood pressure, diabetes and high cholesterol.
While these measurements can be analyzed through visual inspection, AI is well suited to interpreting this kind of information.
“(Electrocardiograms) are almost perfect fodder for computer vision and machine learning,” said Dr. Joshua Landy, Figure 1 co-founder, in comments made to Fast Company. “They’re self-similar, they’re stereotypic, they’re immediately recognizable by an algorithm.”
Even a first iteration of Figure 1’s machine-learning algorithm for detecting ECG images attained approximately an 85 percent accuracy rate. The company’s extensive library of electrocardiograms and other medical images provides a use resource to further train its algorithms and improve their accuracy.
The electrocardiogram computer vision system is just one development in a growing trend of using AI technologies such as object recognition to process medical information. For example, Google this year said its GoogleNet AI object recognition technology successfully detected malignant tumors in breast cancer images 89 percent of the time, compared to an average of 73 percent for human experts.
In another example, researchers from the National Human Genome Research Institute (NHGRI) are using object-recognition technology to identify signs of the rare genetic disease DiGeorge syndrome by analyzing photos of the faces of children potentially suffering from the affliction.
Figure 1, based in Toronto, Ontario, bills itself as the world’s largest network of medical professionals with more than 1 million members. Users can post images of conditions and get feedback from experts all over the world. With its focus on imagery, Figure 1 is sometimes called the “Instagram for Doctors.”
With its large database, Figure 1 could utilize its machine-learning technology to detect and analyze other conditions. For example, the system could examine dermatological conditions or examine the status of a healing wound.
Stephan Cunningham is vice president, product management at Veritone. Working in concert with core internal teams including industry-specific general managers and engineering as well as directly with clients and prospects, he leads the disciplines and business processes which govern the Veritone Platform.