As new technologies shift from early adoption to mainstream, it’s common to find public discourse riddled with inaccuracies and falsehoods. From believing the “cloud” is in the sky to microwaved food becoming radioactive, a curse – and benefit – of being a trailblazer often means educating along with innovating. At Aidoc, we’re okay with that, but it’s time to put these three myths about healthcare AI to rest for good.
Over 500 devices are categorized as “artificial intelligence and machine learning enabled medical devices” on the FDA website. This may create the perception that they are all the same, but distinction must be paid to what the technology does, for instance intelligent decision making or performing a function within a device. Understanding these differences is akin to thinking you purchased a smartphone when in reality you bought an app.
Clinical AI provides augmented intelligence (human + AI) that can help triage suspected findings, improve report turnaround times and streamline care coordination for a radiologist. In all cases, a human is still at the center of clinical decision making. As Nina Kottler, MD, MS, Associate Chief Medical Officer, Clinical AI at Radiology Partners said, “the radiologist will never be replaced, but the role will change and evolve alongside technology.”
Hundreds of elite athletes train for the Olympics, but in each event, only three will make it to the podium. Was it the training or day-of performance that got them there? In many ways, AI is the same. An algorithm can train on a diverse set of annotated data and studies, but if it doesn’t perform accurately “in the wild,” it’s not adding value. Clinical outcomes tied to facility strategy – not the number of data sets used to train – is a far better barometer of algorithm quality.
Navigating the ever-evolving world of healthcare AI can be challenging. Our education and resource hub provides the information needed to deepen understanding about AI and advocate for advancing AI strategy at your facility.