Pioneering healthcare AI with wearables monitoring in cardiovascular patients
There is a clear need for artificial intelligence in healthcare. Introducing AI in the healthtech space, it’s worth noting, will yield solutions for all parties, from providers to patients — making it easier for practitioners to discover medical issues, and reducing the overall cost of treatment for the general public.
The future of AI in healthcare
Today many hospitals and clinics rely on AI to complete their daily tasks. Some leverage artificial intelligence to schedule patients’ appointments more efficiently, facilitate online check-ins, and digitise medical records. Appointment and immunisation reminders, along with warnings involving potential side effects when prescribing multidrug combinations, are also prevalent. “But the most established scientific role in AI so far is in the field of radiology and imaging,” explains cardiologist and event speaker Maria Marketou, “where AI can detect and analyse details that cannot be detected by the human eye.”
Hellenic J Cardiol 2021 Apr 22. Epub 2021 Apr 22. Objective: The efficacy of renal sympathetic denervation (RDN) has…
AI, in most cases, offers an accurate, computed-assisted diagnosis. However, according to Marketou, there are several limitations that must be addressed before innovators can define the future of AI. She emphasises that both human disease and physician experience are hyper-variable, making patient outcomes largely unpredictable where cardiovascular outcomes are concerned. “But the future of AI will be bright,” Marketou adds, “under the precondition that it needs to find what is working and what is not working [in the medical space].”
Stakeholders would do well to make this a priority, as AI could empower providers to interpret billions of data points at unprecedented speed. By reducing human error, defining patterns, and automatically qualifying diseases and patients, it can help to speed up diagnosis and treatment. “AI automation can allow physicians to spend less time on desk work and more time engaging with patients,” explains Marketou. “Of course, it’s available 24 hours — and finally, AI can reduce the cost of care.”
Ultimately, AI is slated to transform healthcare in terms of diagnosis, prognosis, treatment, and even the follow-up plan — yet artificial intelligence systems are only as good as the quantity and quality of their data. “I’m convinced that AI networks are very powerful,” Marketou goes on, “but in order to succeed, we need to keep in mind that they will never be powerful enough to understand the complexity of medicine.” The goal is to build a prediction model that will tell stakeholders what will happen, providing critical information to minimise adverse outcomes and unnecessary treatments.
Developing a framework for wearables monitoring
When it comes to AI in healthtech, remote healthcare monitoring is crucial — especially in the wake of the COVID-19 pandemic, where it became increasingly important to deliver care outside of a clinical setting. “A lot of people think, Okay, this wearable is a step tracker, and I use that to get some step data and some GPS data,” says Jörn Watzke, Senior Director of Global Business Development and Sales at Garmin. “But meanwhile, the wearables changed into a complete wellness tool to help monitor health and also increase awareness.”
Jörn Watzke - Senior Director Garmin Health Global Business Development & Sales - Garmin | LinkedIn
Business development - consumer electronics / services /automotive Product Definition / Portfolio ManagementProduct…
Although Garmin wearables are consumer goods, not medical devices, they are often used for risk assessment or disease prevention. There is a range of options depending on the use cases. By measuring people’s sleep, calories, heart rate variability, activity details, and even oxygen levels and overall stress, experts can leverage parameters that are undeniably pertinent to a person’s health status. As such, wearables are not offering direct diagnoses or perform any treatment based on the data they collect. That means interpreting the information and ordering a medical intervention still lies in the hands of a dedicated physician.
Consider the remote monitoring of someone with depression, for example. A state of chronic depression will change the patient’s sleep patterns and phases, and reveal at least some degree of data variability — for instance, different weekday and weekend stress levels or varying daytime and evening stress levels. Experts can also explore the variability of step patterns and other physical activity data points via wearables, including potential changes in heart rate. From there, practitioners can focus on early interventions, as wearables monitoring allows them to make faster decisions.
The same can be said for pre-diabetes and diabetes, respectively. When a patient enters a pre-diabetes phase without medication, they experience a longer deep sleep phase. Doctors, by recognising these patterns via remote monitoring, can look at the data (available 24/7), and tailor their interventions accordingly. This falls closely in line with Marketou’s cardiology work.
In summary, digital health solutions have increasingly become the norm — with unparalleled movement toward telemedicine largely due to the COVID-19 pandemic. With telemedicine and with wearables monitoring, it’s become easier than ever for patients to access care. Physicians, meanwhile, have access to real-time and historical patient data, enabling them to achieve seamless pre- and post-surgery monitoring. This is a huge driver in heart surgery in particular, where cardiologists and other clinicians can access data quickly to improve patient outcomes.
Bridging the gaps in digital health
Fadi Haddad, Head of Global Business Development at Vienna-based digital health startup Medicus AI, leveraged AI to bring patients clarity where their data is concerned. The initial idea behind the platform was to interpret medical data, making the information less overwhelming for patients and easier to explain for doctors.
“We looked into the current experience, which is giving cryptic information to patients and expecting them to either look up what that information means online or go to a doctor and ask for an explanation,” Haddad explains. “The first challenge we faced was the interoperability of medical data. We developed our own AI-based interoperability engine that uses natural language processing. We use optical character recognition, and we use machine learning in order for us to do matching between the different biomarker names and ranges from different lab information systems and hospital information systems.”
Fadi Haddad - Head of Global Business Development - Medicus AI | LinkedIn
My focus is on bringing new technologies that can solve real problems in healthcare. I joined Medicus AI in 2019 to…
The startup set out to connect diagnostic labs with telemedicine providers, and create specific user experiences for different cases — which would essentially allow patients to bypass their in-person visits (opting instead for a complete remote care approach). Consider this a blend of cutting-edge convenience and striking accuracy.
While appealing in this way, the move toward AI-empowered healthcare isn’t without its challenges. One of the main challenges, according to Haddad, is the interoperability of health data. “If you establish a health data highway, that enables a hospital or lab or startup to plug in — based on the user’s consent, of course — and offer their services in a very easy way.” The idea here is for startups to serve entire countries or demographics.
The second challenge, where digital health solutions are concerned, is reimbursement. “If we get [these solutions] reimbursed, like what is happening in Germany at the moment, we need to convince the doctor to prescribe them,” says Haddad. “So there is a lot of change management and mindset management for the doctor — and these decisions may take generations to change.”
The proverbial gap, however, can be bridged — but first, healthtech innovators must ensure they can trust the data they are relying on to craft their healthcare AI tools.
Can users trust remote systems’ data?
Cardiovascular diseases are the number-one cause of death worldwide — taking an estimated 17.9 million lives each year. IntellIoT, one of six research and innovation actions launched by the European Commission in 2020, aims to develop a framework, appliable to wearables, measuring the vital parameters of cardiovascular patients.
13 international partners involved in the project are working on utilising modern technologies in a solution, that recognises a casualty and informs a physician when intervention is needed. In this way, remote rehabilitation, supported by cutting-edge sensors and next-generation IoT, can significantly decrease negative outcomes.
Anca Bucur, Senior Scientist at Philips Research and the Use Case Lead for Healthcare in the IntellIoT project, delves deeper into the importance of trusting the data on which AI systems rely. To ensure quality and transparency, the IntellIoT consortium has defined three pillars:
- Collaborative IoT (which involves leveraging a range of devices and sensors to support semi-automatic interventions)
- Human-in-the-loop (which guarantees the safety of the system in question and covers the boundary conditions)
- Trustworthiness (with a focus on security and data privacy)
These pillars are paramount to both the research that must be conducted on AI systems like those mentioned above, and to providing access to sufficient data. “Another research aspect here is identifying predictors,” Bucur elaborates. “For instance, what kind of variables would have a predictive value in our environment for either positive or negative outcomes — for either patients that are going to do better or worse.” Here again, high-quality, labelled data is critical. This should include a series of ground rules and outcomes that experts can use for model development.
“The other aspects,” Bucur continues, “are related to having a strong feedback loop and real-world evidence — again, data — so that we actually build trust and differentiate between solutions that are high-quality and actually help, and solutions that are not actually helping.”
Anca Bucur - Senior Researcher - Philips Research | LinkedIn
Anca is currently Product Owner for the Next Gen AI topic supporting the Philips Data & AI Centre of Excellence to…
The overarching goal of devices such as wearables for cardiovascular disease is to reveal that a quasi-continuous, remote support system can provide valid recommendations, effective support, and real clinical benefits. This, of course, must be achieved in a safe way for the patient, backed by ample data.
Join the healthtech revolution
The need for AI in healthtech is clear, and the obstacles, experts like Watzke explain, are less in the technology and more in the collective mindset. This is precisely why innovators like you must participate in the conversation — helping to pave the way forward for remote patient monitoring, wearables, and other digital health solutions that rely on AI.
IntellIoT invites you to join our community and get informed about open calls and future industry events — meetups that bring stakeholders together to discuss artificial intelligence, machine learning, automation, and IoT. To continue the discussion and spearhead the latest innovations, please consider signing up for IntellIoT’s newsletter and spreading the word about the project’s open calls.