by Ken Dropiewski, Principal Consultant – McDermott &Bull Executive Search
Executive Summary: This white paper discussion centers on ways in which three distinct branches of healthcare are currently being impacted by the introduction or use of artificial intelligence (AI). I’ll take a look, via case studies, at various ways AI is being used to assist in improving health outcomes and healthcare in each area. The increased use of AI in healthcare has created tremendous excitement encouraging more spending and the first demonstrations show that deep learning can perform at least as well as human clinicians in some diagnostics. AI’s use in medical devices is a growing trend, as patients and the community at large turn to point of care technology for answers. Finally, perhaps AI’s biggest role in healthcare is its ability to impact disease prevention.
Using AI to Improve Diagnostics in Healthcare
Recently developed AI systems have been used to diagnose heart disease more rapidly and with greater accuracy than human doctors. If widely adopted, this technology has the potential to save billions of dollars and impact countless lives.
Case Study: Ultromics
Most physicians are excellent at their jobs, but none are infallible. Researchers from the John Radcliffe Hospital in Oxford, England, recently unveiled statistics from a proprietary AI developed by Ultromics, a diagnostics system that is proving more accurate than doctors at diagnosing certain types of heart disease.
Cardiovascular disease (CVD) is the leading cause of death globally, affecting 50 percent of the population over the age of 40.(1) Echocardiography is the least invasive way to diagnose heart disease and the most widely used.
However, confirming a diagnosis of CVD by echocardiogram traditionally relies on qualitative judgments by experienced physicians, which may vary according to the data available from a scan.
Ultromics has aggregated data from thousands of scans using over 80,000 data points from each scan to build a learning database which is used to “train” the AI to recognize more subtle indicators of heart disease. The results show machine learning’s ability to overcome subjectiveness, increasing the yield of objective data and outputting highly diagnostic metrics that reflect more information related to the individual’s cardiac function and disease. According to statistics from Ultromics, the technology has improved the diagnostic accuracy of CVD in patients from 80 percent to greater than 90 percent.(2)
AI to Decrease Diagnostic Waiting Times
AI enabled machine learning programs also facilitate in decreasing the waiting times associated with many diagnostic procedures. While the time-to-procedure may not decrease immediately, the time-to- reading and diagnosis will likely decrease due to AI’s ability to fast-track diagnosis.
Case Study: PathAI
Traditionally, pathologists and cytotechnologists determine a diagnosis of cancer by analyzing cytology, tumor tissue slides using a microscope under a variety of stains – a very time-consuming task. In 2015, a group of researchers from Harvard and Beth Israel Hospital piloted the commercialization of the AI technology they developed with startup PathAi.
Most recently, PathAi has partnered with health-tech giant Phillips to further their ability to diagnose cancer and other diseases.(3) Philips, an early adopter of AI tech in other aspects of healthcare, sets its aim for personalized solutions that streamline healthcare, saving time, money, and lives.
AI: Newest Trends in Medical Delivery Devices
Current emerging AI applications appear to be trending towards medical devices that focus on the management of chronic diseases. Companies are using algorithmic machine learning models to monitor patients via sensors and then automate the treatment delivery at point of care using connected mobile apps.
Case Study: Medtronic Sugar.IQ and MiniMed 670G
For years, Medtronic has provided systems like continuous glucose monitoring and insulin pumps, used predominantly by people on insulin to manage their blood sugar levels. Recently, the focus has shifted to use of big data and AI as a way to utilize their technology and devices to not only provide continuous feedback but ultimately improve quality of life and influence the better outcome.
Diabetes currently affects 30.3 million Americans. (4) Caused by sustaining high levels of blood glucose or blood sugar from eating certain foods, in patients who do not have a normal endocrine response or are lacking in insulin production.
Sugar.IQ uses IBM’s Watson AI to provide three features key to managing patients with chronic diabetes.
1 . Insights – The app continuously gathers data, uncovering patterns and behaviors that are associated with the patient’s glucose levels. The device sends personalized messages delivered in real-time to help the patients and caregivers understand how specific actions or habits may be affecting their glucose levels.
2. Glycemic assist – Users are able to ask the app to follow any specific food- or therapy-related events. Sugar.IQ uses AI databases algorithms to help discover the impact these items have on glucose levels and also enter data that can be used later to predict outcomes.
3. Food logging – By quickly and easily tracking food intake the app can deliver insights that impact an individual’s food choices and glucose levels.
Launched last year, MiniMed 670G system, the first insulin pump technology of this kind receiving full FDA approval. Medtronic describes a system that uses AI to automate the delivery of insulin, monitoring it continuously to stabilize blood glucose levels 24/7. Using machine learning algorithms from the Sugar.IQ app, the MiniMed 670G system is trained to “self-adjust its insulin delivery every few minutes, offering better control of blood sugar levels in at-risk patients.(5)
AI’s Impact on Disease Prevention
“The aim of medicine is to prevent disease and prolong life, the ideal of medicine is to eliminate the need for a physician.” William J. Mayo
With a turn toward value-based medicine, improving outcomes through disease prevention has never been more important. Today’s health-tech companies are leveraging big data, natural language processing, and data-driven algorithms to create AI programs that not only predict disease at the earliest signs but also work to uncover risk factors and prevent disease occurrence altogether.
Case Study: HBI Solutions – Spotlight Risk Model Engine
When you look at HBI’s customer base you will find listed the foremost health delivery organizations, federal health centers, ACOs, insurers, health information exchanges, and a vast network of technology vendors. The data scientists at HBI have successfully leveraged massive amounts of data from their customers, devised algorithms and applied AI to proactively identify at-risk patients, improve outcomes and provide care options at a lower cost. By linking together previously unidentified risk profiles with known outcomes, HBI discovers disease at its earliest and most preventable stage.
The Spotlight Risk Model Engine is the company’s foundation. The engine takes in all relevant datasets, including clinical information from a patient’s EHR, insurance claims, billing, and even other unstructured data, to provide a real-time risk model that will continually update as new data enters the system. The result: using AI, predictive analytics, and performance analysis, HBI offers healthcare the greatest edge in preventative medicine. For example, recent data shows that HBI’s model was able to predict 72.9 percent of chronic kidney disease cases an average of 90 days prior to diagnosis. (6)
The Business of AI in Healthcare
It’s impossible to discuss the role of AI in healthcare without briefly talking about the numbers. Cisco predicts that healthcare spends into IT will continue to surpass the cross-industry average. Electronic health records (EHR) represent a significant data source for AI algorithms grew from 40 percent in 2012 to 67 percent in 2017. Investors consider AI to be the next big thing in healthcare as well. This is evidenced by investments in venture capital toward cutting-edge, medical tech that already uses AI like computer vision, deep learning, machine learning (ML), and robotics. This segment has skyrocketed from just $30 million in 2012 to nearly $900 million in 2016. A further breakdown of healthcare segment buy-in shows largest growth in business, diagnostics, treatment, and surgery.
In Summary
Even if William J. Mayo’s version of healthcare, where doctors are no longer needed, isn’t right around the bend, it seems to have people talking. With AI-based tech, we are making strides toward improving patient outcomes, preventing disease through better diagnostics, and predictive risk models, followed by behavior and lifestyle changes. AI’s ability to leverage massive amounts of data, with known outcomes- based medicine, and societal trends through technology that is becoming less expensive to develop, ensure its continued impact for the foreseeable future.
(1)http://www.who.int/mediacentre/factsheets/fs317/en/
(2)http://www.ultromics.com/technology/
(4)https://www.cdc.gov/media/releases/2017/p0718-diabetes-report.html
(5)http://www.healthcareitnews.com/news/medtronic-introduces-ibm-watson-powered-sugariq-diabetes-app
(6)https://medinform.jmir.org/2017/3/e21/
Ken Dropiewski has been in the med tech space for over 20 years. He is a Principal Consultant with McDermott & Bull Executive Search, recognized by Forbes as one of the industry’s best with offices across North America and in Europe. Ken is hired by innovative device and biotech companies to help build their best teams possible at the executive level.
