Your Resource for All Things Apps, Ops, and Infrastructure

Gimme Five: The Promise of AI in Healthcare

Artificial intelligence has vast potential to drastically transform patient care, early disease detection, and hospital efficiency. Healthcare IT leaders are embracing AI, with some already applying it and many more exploring its capabilities. In fact, 75% of healthcare organizations see an AI strategy in their future. 

We sat down with AHEAD Field CTO and AI specialist Josh Perkins to get insights into the current use cases, top challenges, and future possibilities of AI for healthcare. 

There are many potential use cases for AI within healthcare. What do you see as the top opportunities available to healthcare systems today in terms of patient care? 

Josh: One of the most valuable patient care use cases for AI is around imaging, specifically within radiology, oncology, and pathology. This comes down to the fact that a computer system has the ability to evaluate and analyze more scans and medical images than any one practitioner. AI can analyze those images and recognize variances on a scale way beyond human capability. Imagine looking at 1,000 images simultaneously and being asked to pick the one with the minute anomaly—AI can do that. The result of this is that practitioners can achieve far earlier detection of illness, and increased accuracy of diagnosis as well. 

What are the opportunities around healthcare system operations? 

Josh: There are many operational benefits to hospital systems in terms of gaining efficiency through analyzing patterns. One of the most common use cases currently is for nurse scheduling. With AI, hospitals can understand the optimal number of nurses needed on any floor at any given time.  

There are also disease-tracking implications, some that we’re seeing in use right now with COVID-19. Through data input and analysis, AI can help us see the way a disease spreads within a metropolitan area and which healthcare systems will need extra preparation and equipment. In this sense, AI can save lives because we can get an early warning as to the expected impact and pressure on certain metropolitan areas. 

Another use case is billing practices. Healthcare systems can run their business more efficiently by using AI to set expectations around, for example, how long it will take to go from treating a patient to seeing payment from their insurance company. This is really important to keeping a hospital operational and cash-flow positive. 

What challenges exist around adopting AI? What are the concerns we must overcome before AI is adopted on a large scale within medicine? 

Josh: The biggest challenges around AI in healthcare right now are, at a high level, around maturity and trust. The idea being: how do we evolve towards being comfortable with AI being part of the diagnostic process within medicine in general? 

On a more practical level, it’s about data. In order to fuel AI, we have to feed in a lot of information. The more data seeded, in theory, the better algorithms are at finding insights. The problem here is that a lot of the data within the healthcare industry has a real security or governance concern, let alone regulatory considerations. For hospitals, especially research hospitals, one of the challenges is taking data sets and making them viable for use in research while maintaining regulatory compliance and safeguarding patient privacy. 

What does the future of AI in healthcare look like in your opinion? 

Josh: I think there will be giant strides connecting AI to the Internet of Things (IoT) within medicine. Think about a hospital room—there are many apparatuses connected to or near the patient. In the future, all of these devices will output data which will be funneled into one place and analyzed by AI algorithms. AI will perform real-time analysis of the patient to help determine treatment or predict health events that are on the horizon. 

An example of this would be a patient at high risk for a heart attack. Over time with a vast collection of data from previous patients, we’ll be able to see the precursors of a coronary event on an EKG before it even happens. This can also apply to stroke care or cancer care where doctors can practice intervention before the negative health event, whereas before we were only able to acknowledge the existence of an event. 

Speaking of IoT, what part do you think wearables will play? 

Josh: From a clinical engineering perspective, we’ll have those on-premise IoT devices that are feeding in sensor data. But, outside of healthcare facilities, remote IoT devices will also be data producers, broadly categorized as wearables. This isn’t things like Fitbit or Apple Watches, but devices built specifically to collect data around a certain health condition. For example, Avery Dennison is a film and packing materials manufacturer. They have a bioengineering product that’s literally a peel-and-stick heart tracking sensor. It sticks to the body for seven days and records all of the wearer’s cardiac activity over that time period. Their doctor then takes that information and uploads it into a system for analysis. We’re seeing technology like that to some extent, but that’s really going to expand with the rollout of 5G as more devices become connected devices.  

What should enterprise IT leaders be asking around the implementation of AI? 

Josh: One of the biggest shifts that will take place concerning AI is that IT leaders will need to think outside of the infrastructure and delivery layer. In order to respond to the impacts of AI, they have to become more familiar with and aware of the actual datasets they’re using and the implications of that data traversing different systems. It’s really a security issue that will require immense stitching between different technology areas. 

Healthcare systems will also have to build a base before they can truly begin taking advantage of AI on a large scale. A lot of these algorithms will be deployed from the cloud which will still require all the foundational elements of cloud utilization like security, compliance, governance, and networking. Those elements will all have to be in place and well established before the institution can leverage forward-thinking platform services that get them to those AI algorithms. 

Subscribe to the AHEAD i/o Newsletter