Former Philips Imaging CEO heading machine learning startup to improve stroke diagnosis

The critical area of a brain scan (in red), as determined by MedyMatch software--Courtesy of MedyMatch

Former Philips Imaging Systems CEO Gene Saragnese has a new gig. He's seeking to deploy machine learning software to help radiologists diagnose stroke as the head of new company MedyMatch Technology.

The 20-person Israeli company just exited stealth mode. It announced the opening its first North American office in Boston, where it will seek to commercialize of its cloud-based software aimed at U.S. hospitals.

Saragnese said in an interview he's aiming for FDA approval this year, via a to-be-initiated retrospective clinical trial to test the accuracy of its technology. Apparently, by extracting information from complex scans of the brain, the MedyMatch software can correctly diagnose the occurrence and type of stroke--ischemic or hemorrhagic--with 95% accuracy, compared to around 70% using the human eye alone.

"In this particular case, most of the legal cases are filed because patients feel they did not get the chance to receive treatment," Saragnese said.

MedyMatch's technology is designed to analyze the images, and send them back to the hospital via the cloud, but with the crucial areas highlighted. The goal is to help the average radiologist perform like an expert, Saragnese said.

Like everyone else, stroke sufferers are generally sent to whichever hospital is closest, and they aren't usually very experienced at responding to stroke, Saragnese said. Most medical imaging errors occur because of the inability of the radiologist to perceive subtle aspects of the scan, according to the CEO.

Studies demonstrating the effectiveness of stent retrieval devices to treat stroke by physically removing the clot from the brain have brought about an ongoing revolution in the treatment of stroke, but they are only useful for treating ischemic stroke, making the course of clinical action very dependent on that initial diagnosis.

The typical hospital doesn't have the devices. Under the emerging treatment paradigm, patients who are eligible for a stent retrieval device are supposed to be quickly shipped to a specialized stroke center. "That's going to open up a new window for treatment, but you've got to get the patients to those specialized hospitals first," Saragnese said.

He also said that speed is a crucial component of clinical success. Guidelines from the American Stroke Association recommend initiating treatment with the devices within 6 hours of symptom onset, and say their effectiveness is uncertain beyond that point.

Saragnese believes MedyMatch software can prevent chronic disability and save lives, as well as money. From 800,000 strokes in 2012, the American Stroke Association estimates there will be 3.4 million instances by 2030 at a cost to the healthcare system of $240 billion (up from $105 billion in 2012).

Meanwhile the incidence of stroke in China is supposed to rise from 2.5 million cases per year to 4.4 million in 2030, Saragnese said.

And he believes that the software has potential applications in the diagnosis of pulmonary embolisms and head trauma as well. "We are in the business of providing decision support tools that are derived from artificial intelligence," the CEO said.

MedyMatch lists Jerusalem's Hadassah Medical Center, Massachusetts General Hospital and Capital Health hospital systems as its clinical partners.

The CEO said funding comes from hedge funds and angel investors. MedyMatch's board of directors includes individuals from Genesis Capital Advisors, Ziegler Meditech Equity Partners and Exigent Alternative Capital.

Meanwhile, IBM's Watson Health unit is also seeking to use its machine learning capabilities to read medical images, as evidenced by its billion-dollar acquisition of Merge Healthcare. The company is developing software, dubbed Avicenna, that's supposed to integrate information from medical records into its analysis.

- read the release