San Antonio — Algorithms are now a silent, ever-present part of life, turning the wheels that operate our favorite apps, games, television programs, and more.
As the complexity of those equations expands, artificial intelligence and deep-learning algorithms are increasingly extending into fields such as healthcare, including diagnostics, which aim to catch health problems in patients. IBM’s Watson has most famously brought attention to the potential AI systems have for healthcare, but startups are helping artificial intelligence spread into specific sectors of the industry, too, such as radiology.
One radiology contractor, Intrinsic Imaging, which is based in both Boston and San Antonio, TX, says it has seen an uptick in recent years of the number of clients that want help assessing the accuracy of AI software used to diagnose conditions such as cancer in X-rays, CT scans, and other healthcare imaging. Intrinsic has been hired to work on 14 clinical trials of medical devices that use a form of AI software since Intrinsic was founded in 2010, and about 10 of those have come in about the past two years, according to Todd Joron, president and chief operating officer.
Devices in two of those trials have already received FDA approval, while the clinical work and regulatory reviews of the others are ongoing, Joron says. (As a side note, the FDA has been studying how it plans to regulate the use of artificial intelligence in its regulatory approval process.) Another unnamed company picked Intrinsic this week to conduct an upcoming clinical trial that will study how effective its software is at detecting diseases in the musculoskeletal system.
“The application of this type of technology, because it’s a visual science, is so directly applicable to radiology that people are investigating all different types of therapeutic indications,” Joron says. “The actual software has gotten to a level that it’s sophisticated enough to be effective, so many companies are jumping on that bandwagon. We’re seeing it because we’re better known now, but there’s also a huge rush going on.”
Intrinsic declined to reveal the names of the companies it works with, except for one: Nashua, NH-based iCad, which has a product that uses computer-assisted detection to find cancers. ICad also has a treatment for some early stage cancers.
Plenty of other startups have entered into the space globally, including Andover, MA-based MedyMatch, which uses AI software to classify images; Israel-based Zebra Medical Vision, a company with a large database of medical images that aims to help radiologists detect overlooked indications, among other things; and Seoul-based Lunit, which is developing technology to help physicians make clinical decisions, not just diagnoses.
The increased use of AI and deep learning algorithms in medical startups is trickling down to contractors like Intrinsic Imaging, which was founded to be a core lab that helps medical device companies design and run the imaging component of their clinical trials. Images of X-rays and CT scans (among others) are used to evaluate the impact a device might be having on a health condition, and Intrinsic designs the protocols for how clinical testing sites should make the images.
Intrinsic also determines what parameters should be measured to show how the device is performing, and its staff of radiologists will read and interpret the images throughout the life of the trials, Joron says. Intrinsic has 70 full-time radiologists, and contracts with 100 others, who read the scans. The company also does work on clinical trials for therapeutics.
Intrinsic’s radiologists read the images twice when the company is working with AI-software makers: once on their own and a second time with the AI program helping out. The sponsor of the trial will then compare the difference from the radiologists’ readings with and without the AI algorithms.
“There’s a huge boom in the market going on and we’re benefitting tremendously,” Joron says.
AI is a hype-heavy sector of tech, so it will be important to watch how many of these emerging medical-technology companies are able to remain afloat.
Notably, Joron isn’t conflicted about helping artificial intelligence startups that might want to replace radiologists with algorithms because of the level of education it requires and the “cerebral” portion of the job.
“In the near-term, these AI products will get very good and elaborate—they’ve improved immensely over the last 10 years,” Joron says. “I would say that the radiologist’s job is pretty secure.”
