Artificial intelligence ramped up in 2016, but has to prove its utility in 2017

Image: Pixabay
While a lot of buzz surrounded artificial intelligence in 2016, helped along by IBM's dealmaking rampage around its Watson supercomputer, players still need to figure out where best to deploy these technologies and how to create a business around them.

According to MedyMatch Chief Financial Officer Michael Rosenberg, 2016 was a banner year for artificial intelligence. How does he know? “Even my mother knows what it is now. It’s going mainstream.”

But while everyone, it seems, is bandying a bunch of buzzwords–artificial intelligence, machine learning, cognitive computing–the technology is still finding its footing, with companies working out how, exactly, to deploy AI in healthcare, determining which applications it is most useful for and how to build a sustainable business model around it.

A proliferation of players, from IBM and GE-backed Arterys to Houston Methodist Hospital and Harvard University, have sought to apply artificial intelligence to various areas. And while Rosenberg expects AI to be able to "optimize almost every area" in healthcare, he notes that it can be divided into two distinct approaches. The first is preventive medicine and population health, where AI can be used to comb through medical records to find abnormalities or patterns that can be flagged for a physician. The second is as a decision-making tool, which a physician uses to help direct treatment in real time. The value here, Rosenberg said, is giving the physician a second opportunity to make the best decision.

IBM, which broke out its Watson Health as a separate unit last year, is working on both sides. In October, it teamed up with Siemens Healthineers to apply its IBM Watson Care Manager to bring value-based care to patients with chronic conditions. This particular iteration of Watson is designed to help patients and caregivers collaborate to support individual health, the company said. IBM has also created versions of Watson for use in areas such as oncology and drug discovery and embarked on numerous partnerships to put its cognitive computing capabilities to work. In a partnership with Pfizer, Watson will analyze vast amounts of medical literature and clinical data to help the pharma accelerate its immuno-oncology research. And the Department of Veterans Affairs is using Watson for Genomics to scale up its cancer precision medicine program.

Meanwhile, imaging looks to be a hotspot for AI. Last year, we highlighted Arterys, which is working on software that reduces the time it takes to interpret cardiac MRI scans. This summer, MIT and Harvard scientists developed a system that analyzes breast cancer imaging with a machine-learning algorithm and Houston Methodist researchers unveiled software that rapidly and “intelligently” reviews pathology reports and mammograms to determine breast cancer risk.

MedyMatch, helmed by former Philips Imaging CEO Gene Saragnese, teamed up with Capital Health to create an AI-based decision support system to diagnose stroke patients in the emergency room. The tool will analyze medical images and provide an ER radiologist with information to rapidly make treatment decisions.

And while there is a lot of “noise” surrounding the technology, getting AI off the ground relies on proving how it can be used to improve healthcare. IBM is conducting concordance studies to assess the level of agreement between human doctors and Watson; the company recently reported study results that IBM for Oncology agreed with doctors 80% of the time on treatment decisions for patients with nonmetastatic breast cancer. But IBM doesn’t see artificial intelligence replacing doctors. Where the real value lies, said Watson Health Deputy Chief Health Officer Dr. Andrew Norden, is taking the intellectual burden of reading and analyzing literature off the physician, who can then spend that time interacting with the patient.

That said, Rosenberg can see “rapid adoption” of artificial intelligence in areas where the doctor-to-patient ratio is low. While such technologies wouldn’t be the equivalent of physicians, deploying them in rural areas could offer a level of assessment or expertise that would otherwise not be available, he said.