One of These Things Is Not Like the Other:
Sheepdog or mop?
Puppies or food?
Carcinoma or keratosis?
Big data, artificial intelligence and machine learning were front and center during HMS Momentum 2019, in which experts from across the healthcare spectrum came together to discuss the future of the industry. And though we’re still reeling from three days of eye-opening discussion, collaboration and perspective, one clear takeaway is that these technologies are transforming healthcare as we know it — and that comparing images of labradoodles and fried chicken can actually help put artificial intelligence (AI) into a healthcare context.
Human vs. Machine
AI’s unmatched ability to ascertain the normal from the abnormal has opened the door to groundbreaking advancements in disease diagnosis. It’s led to the development of IDx-DR, for example, a device that can immediately and autonomously detect diabetic retinopathy and the first AI diagnostic to receive FDA approval.
Machine learning systems and deep neural networks continuously learn from the information they receive, progressively building upon this knowledge to get better and better at discerning real from fake — benign from malignant — and so on.
When it comes to AI, humans have always had a bit of a leg-up on the machines in that there’s no complex algorithm behind the fact that we know a dog when we see it. It’s part of what makes the “dog or” comparisons amusing — it might be a little hard to tell them apart at first, especially when part of a grid or series of images, but generally speaking, we would never actually confuse a puppy for a food item. Or would we?
Well, we might. And when it comes to highly complex diagnostic images, it’s the speed and scale at which we’re able to make these distinctions where the technology has us beat.
A Promising Use Case
In his keynote address at HMS Momentum, Dr. Joel Selanikio, award-winning physician, health and technology activist and CEO of the data collection, messaging and visualization company Magpi, helped us understand how exactly this concept applies in healthcare. He cited a recent study published in the journal Nature in which a deep convolutional neural network (CNN) was trained in 129,450 clinical images of skin disease. The CNN learned to detect the most common and deadliest skin cancers with a level of competency comparable to, or better than, the 21 board-certified dermatologists against whom it was tested.
A key takeaway from Dr. Selanikio’s presentation and a common thread throughout Momentum was the role of big data, analytics and AI in making healthcare accessible to a greater number of people — and the recognition of technology as an equalizer rather than a divider. The commercialization of IDx-DR is an example of this — by making these technologies more widely available to providers and potentially even consumers, diagnoses and treatments not only become more precise, but also less costly and time-consuming. And in the ongoing effort to deliver higher quality care at a lower cost, AI’s potential in improving healthcare is profound.
The Future Is Now
At HMS, we’ve implemented AI and machine learning across our technology and solutions to enhance workflows, while continuously improving upon these systems to ensure maximum results for our healthcare clients. But more so, we are helping to structure the conversation around how technology can drive positive, measurable outcomes across your patient and member populations.
What AI applications are you most excited about?