In the last edition of eFYI, the Tech Savvy column took a look at The Internet of Things, commonly referred to as IoT. Hospitals are starting to use IoT in three areas: inventory management, workflow optimization and device integration.
Inventory management will be useful in areas like pharmacies and overall inventory control in warehouses, while workflow monitoring using radio frequency identification (RFID) in badges and tags will allow hospitals to identify and address bottlenecks.
Device integration with IoT has more of a consumer focus, as people look for ways to integrate devices like Fitbits and others to bring patient-provided data into the fold of care delivery. Glucometers, blood pressure cuffs, smart insulin pens – basically any device that can collect data and statistics of patients -- may be useful for data integration. The idea is that with more complete patient data available, faster and more accurate healthcare decisions can be made or verified.
This will create a lot of data, and it begs the question: Who will have time to analyze it all? This is where we turn to computers to help identify which data is important to which patient and what it is saying about the patient’s health.
Enter Artificial Intelligence, or AI
AI is a branch of computer science that aims to create intelligent machines that can one day learn without any kind of supervision. In 2012, Google announced that one of its AI computers (after watching thousands of hours of YouTube videos) had trained itself to identify cats! The point is: the computer had not been programmed to identify cats, but it learned how to do so without direct supervision.
AI is only in its infancy
There are different levels of AI of course, and we are only at the infancy of this technology. However, you may already benefit from AI. For example, when you ask your smartphone for nearby restaurants, your smartphone uses AI to not only list the ones closest to you, but to also present the ones it predicts you might like the most based on your restaurant history.
AI also comes into play when your robotic vacuum cleaner analyzes your home layout and develops a strategy to clean your floors in the most efficient path possible.
What AI has to do with diabetes
An area in diabetes care that AI is targeting is diabetes-related retinopathy (DR), which can largely be prevented if caught early.
The problem is that there are not enough qualified ophthalmologists, worldwide, who can make this diagnosis.
AI as a possible solution
The California Health Care Foundation (CHCF) produced great results in 2015 using its own deep learning algorithm to detect DR. This algorithm agreed with doctors 85% of the time (doctors agree with each other about 84% of the time on DR diagnosis).
More recently, Google Brain, the AI research team at Google, has worked with doctors in India and the US to help them diagnose DR. They collected over 128,000 images which were then evaluated by three to seven ophthalmologists from a panel of 54. Using a deep learning algorithm and these images, they created an AI model to detect DR. They tested their model on two data sets totaling 12,000 images.
The predictions from Google Brains AI algorithm were so close that it is considered to be on par with the diagnosis of experienced ophthalmologists. You can read Google’s paper at The JAMA Network.
A future of limitless possibilities
If AI systems continue to prove themselves and clear regulatory challenges, it’s possible that millions of people will benefit from early detection and treatment because AI systems can analyze images 24 hours a day, 7 days a week.
Artificial Intelligence will benefit strategic healthcare areas at first, but as our ability to create more intelligent AI systems grows, the impact of AI will also grow in all areas of health care.