WWW.MEDTECHSTRATEGIST.COM

VALUE ASSESSMENT TOOLS

In Medtech, Real-World Evidence is Moving Beyond Theory

to Regulatory Use Wendy Diller, 16

EXECUTIVE INTERVIEW

Managing in Interesting Times: An Interview with Paul LaViolette

David Cassak, 6

DIGITAL HEALTH STRATEGIES

ResMed: Using Digital Health to Fuel a

Medtech Growth Strategy Mary Stuart, 20

MARKETING STRATEGIES

Digital Innovation: Consumer Activation Comes

of Age in Medtech Ajay Gupta, Karen Passmore, Sundar Ganapathy, and Chris Smith, McKinsey & Co., 30

EXECUTIVE INTERVIEW

What’s Next for the Ischemic Stroke Market?

Brett Wall, SVP and President of Brain Therapies at Medtronic, Weighs In

Mary Thompson, 38

START-UPS TO WATCHFocus on AI and Screening

Ophthalmology

Eyenuk: AI Boosts Diabetic Retinopathy

Screening RatesMary Stuart, 48

Cardiology

HeartSciences: AI Improves Frontline Risk Stratification for

Heart Diseases Mary Stuart, 52

JUNE 8, 2018Vol. 5, No. 8

MedTech

STRATEGIST

MEDTECH STRATEGIST © 2018 Innovation In Medtech, LLC. All rights reserved.

52 START-UPS TO WATCH

Even though the ECG is the standard screening tool used by physicians to detect heart disease, existing ECG technology has limited effectiveness in detecting structural heart disease and coronary artery disease.

While many serious diseases quietly lurk in an asymptomatic phase during which they could be treated to prevent undesirable outcomes, it is rarely cost-effective to screen the general population for their presence, especially when screening involves expensive imaging modalities and interpretation by specialists. The applica-tion of advanced technology including artificial intelligence (AI) to medical processes is chang-ing that paradigm, with its capability of extract-ing new kinds of diagnostic information from

non-invasive, conventional systems that are fairly ubiquitous in the practices of first-line physicians. New kinds of AI-enabled screening systems could offer primary care and other doc-tors an expert opinion that would allow them to make the determination to refer a patient for further care or confidently rule out disease dur-ing a patient visit.

HeartSciences Inc. is working on offering just that proposition in cardiovascular disease. By applying signal processing and machine learn-ing (a type of artificial intelligence) to the signals

gathered by conventional 12-lead electrocardi-ography, HeartSciences can pick up signs of left ventricular early diastolic dysfunction, an early marker of many heart diseases including heart failure with preserved ejection fraction, valve disease, and coronary artery disease (CAD) asso-ciated with diastolic dysfunction.

HeartSciences began in 2008 as Heart Test Laboratories, to improve the capabilities of the resting ECG, which is the frontline tool for heart disease. There has been a lack of innovation

in the ECG segment, apart from making it available in new form factors such as smart-phones, notes CEO Mark Hilz, who says “It seemed to me ridiculous, after 100 years, to be only looking at an analog waveform.” The start-up’s focus was on advancing the clini-cal effectiveness of ECG technology so that it would be a more effective screening tool for patients, in order to reduce the large number of patients with heart disease who remain undiagnosed until an adverse event occurs. Currently, 50% of men and 64% of

women who die suddenly of a heart attack had no prior symptoms.

Even though the ECG is the standard screening tool used by physicians to detect heart disease, existing ECG technology has limited effectiveness in detecting structural heart disease and coro-nary artery disease. For example, current 12-lead resting ECG technology detects CAD less than 50% of the time. The lack of low cost, effective frontline tools for cardiac testing has also led to expensive diagnostic testing with low yields that may put patients at unnecessary risk. A study by

HEARTSCIENCES: Artificial Intelligence Improves Frontline Risk Stratification for Heart Diseases

SOUTHLAKETEXAS

contact

Mark Hilz, CEO mark.hilz@heartsciences.com

year founded2008

who’s behind itAndrew Simpson (Chairman); Mark Hilz (CEO); Aaron Peterson (VP R&D)

unmet clinical needA frontline screening tool suitable for primary care settings to find heart disease in early stages while it is most treatable

solutionThe MyoVista 12-lead resting electro-cardiograph utilizing continuous wavelet transform-based signal processing and artificial intelligence to identify diastolic dysfunction, an early marker of heart disease

funding to date More than $32 million provided by high net worth individuals and family offices

board of directors Andrew Simpson (Chairman); Mark Hilz (CEO); Richard Gray, MD (Tyler Heart Institute); Brian Szymczak (Apollo Endosurgery); John D. Owen (co-founder, JetBlue Airways); John Q. Adams, Sr. (who has founded and sold several pharmaceutical companies including Mucinex)

The current screening paradigm for heart disease is missing people at early stages, when diseases are most treatable. HeartSciences has taken a ubiquitous screening tool, the 12-lead ECG, and has, through the application of signal processing and artificial intelligence, turned it into a powerful tool that can detect diastolic dysfunction, an early indicator of most cardiac diseases.

by MARY STUART

Cardiology

JUNE 8, 2018

53START-UPS TO WATCH

Duke University of 400,000 patients in the United States undergoing invasive catherization testing found that only 38% of patients without previously known heart disease had blockages that required treatment.

From 2008 to 2013, HeartSciences focused on fundamental research and proof-of-concept using signal pro-cessing techniques known as wavelet transform mathematics to transform the ECG’s analog waveform into a larger dataset of wavelets that could be analyzed to capture the energy kinetics of the heart throughout its cycle. In early 2013, the company had some insights and an early prototype called the MyoVista ECG, but ran out of money. Hilz and Chairman Andrew Simpson, both previous investors in Heart Test Laboratories, stepped in to fund the company, continue R&D and move the technology forward to com-mercialization. The company name was changed to HeartSciences.

Neither Hilz nor Simpson had pre-vious medtech experience, but Hilz says, “We are adding technology to the ECG, and I have been in the technology industry for 30 years. Mr. Simpson’s background is in invest-ment banking and running billion-dol-lar operational companies. Those two things combined make us well suited to understand technology and to be able to finance it.” To date, Heart-Sciences has raised more than $32 million from high net worth individu-als and family offices.

Upon taking the helm, the new leadership enlisted the help of Par-tho Sengupta, MD, currently Chief of Cardiology at West Virginia University, Heart and Vascular Institute, as well as the director of Cardiovascular Imag-ing, and now a clinical advisor to the company. Sengupta led a clinical study at the Icahn School of Medicine at Mount Sinai (New York, NY) to provide

further understanding of MyoVista’s capabilities and to determine what fur-ther refinements were needed. A total of 188 patients referred for CT angiog-raphy (who would thus ultimately get a definitive diagnosis) received a base-line 12-lead surface ECG, which was processed by the MyoVista wavelet transform mathematics. Patients also underwent a complete echocardiogra-phy evaluation and CT angiogram.

In analyzing the study results, Sen-gupta found that although MyoVista correctly identified coronary artery disease in 82% of the cases, it could also detect abnormal myocardial relaxation, a sign of an even more impactful harbinger of heart disease: diastolic dysfunction. “That’s impor-tant,” Hilz says, “because diastolic dysfunction is one of the first detect-able effects that heart disease has on the heart tissue, for almost all types of heart disease; including structural, ischemic, and severe arrhythmias. If we can insert this kind of capability

into a regular 12-lead resting ECG, it would significantly improve risk strat-ification.”

The study, which was recently pub-lished in the Journal of the American College of Cardiology, found that abnormal mechanical relaxation was seen in 70% of the patients upon echocardiographic assessment; Myo-Vista predicted abnormal myocardial relaxation with a sensitivity of 80% and a specificity of 84% (“Prediction of Abnormal Myocardial Relaxation from Signal Processed Surface ECG,” Sengupta, Kulkarni et al, JAAC, April 17, 2018).

For the study, MyoVista signal pro-cessing provided 370 data variables for analysis, extracted from the T wave of each patient’s ECG test. (Now the company is analyzing approximately 650 variables, not including the tradi-tional variables, and is also starting to work with QRS parameters; through machine-learning, the platform con-

Online print subscriptions, reprints, and web posting and distribution licenses are available.Contact Bridget Kelly-Stoll | 888-202-5939 | b.stoll@medtechstrategist.com

Source: HeartSciences Inc.

Figure 1

MyoVista Informatics

MEDTECH STRATEGIST © 2018 Innovation In Medtech, LLC. All rights reserved.

54 START-UPS TO WATCH

tinues to evolve). “I give Dr. Sengupta a lot of credit for guiding the com-pany,” says Hilz, “for recognizing that the large multi-variable problem called for artificial intelligence and machine learning.” Because HeartSciences was initially looking at coronary stenosis, Sengupta recommended that each patient in the study also undergo a thorough echocardiography evalua-tion along with CT angiography. “It was Sengupta who realized the impli-cations of picking up abnormal cardiac relaxation, a surrogate for diastolic dysfunction,” Hilz says.

To further train the machine-learn-ing algorithm for the clinical device, Hilz estimates that the company needs to test 1200 patients of diverse ages and health statuses. To that end, HeartSciences has collaborations with Mount Sinai, West Virginia Univer-sity, the Windsor Cardiac Centre in Ontario, Canada, and UCLA (Univer-sity of California Los Angeles). Data has been collected on 900 patients so far, Hilz reports. “Once we get to 1200, we will install the final version of the machine-learning algorithm into the product.”

Upon market clearance (the com-pany plans to submit to the FDA in late 2018), the initial commercial strategy is to prove the clinical effectiveness in the marketplace. The company has received CE Mark and it launched the

product in Europe in late 2017. No marketing partnerships are currently in place, but Hilz notes there is a great breadth of companies that would be logical partners for a technology like

MyoVista, from the ECG manufactur-ers, to cardiovascular device com-panies that don’t currently have a foothold in the ECG market, to infor-mation tech companies like Google.

The commercial version of the prod-uct, for which two utility patents have been issued and others are pending, is a full-featured 12-lead ECG with the addition of the MyoVista wavelet ECG functionality. The user gets a touch screen that displays the conven-tional ECG tracing annotated by the

accepted Glasgow Analysis, as well as additional graphics in which colors represent the energy distribution dur-ing the cardiac cycle and the energy frequency during the heart’s repolar-ization cycle. A specific energy detec-tion algorithm is displayed to indicate diastolic dysfunction (see Figure 1).

Pricing of the MyoVista systems is currently comparable to that of the fully functional ECG devices sold by General Electric Co. or Royal Philips Electronics NV. “It is very intentional to offer a very low cost, improved, front-line tool,” Hilz says, since, as noted, the vast majority of patients with heart disease are not being diagnosed. It’s estimated that left ventricular diastolic dysfunction affects 20-30% of the gen-eral adult population.

At the same time, the goal is to avoid paying for a series of tests—stress tests, stress echo, and combi-nations thereof—for patients who don’t need it. “We need improved risk stratification tools,” says Hilz. “Today, a primary care physician has a regular ECG and a questionnaire. We need to enhance that frontline tool to say who should go on for further testing, and diastolic dysfunction is the right thing to detect. If the system identifies dia-stolic dysfunction while patients are still asymptomatic, we could treat them sooner, before they have a heart attack or develop heart failure.”

“It is very intentional to offer a very low

cost, improved, frontline tool,” Hilz

says, since, as noted, the vast majority of patients with heart

disease are not being diagnosed.

CONNECT WITH US!