6 things that could help advance AI adoption in radiology
Artificial Intelligence (AI) is at an inflection point, and radiology will need to address six critical components to move into the mainstream.
Radiology has always been on the cutting edge of technology in medicine, and artificial intelligence (AI) has made some progress within this specialty. But AI is at an inflection point. To make use of AI as a more accessible, trusted reality, radiology will need six critical components:
1) Reassurance that AI will never replace a physician
With the growing body of evidence that shows the extraordinary potential of AI in healthcare, I find it interesting that widespread adoption remains slow. Although I hear multiple reasons for the reluctance of radiologists to embrace AI, one of the most common that I hear is the misperception that AI is going to put us radiologists out of business. This simply isn’t realistic. At the end of the day, algorithms address technology; radiologists take care of people, with all the inherent responsibilities, nuances and complexities. Algorithms aren’t designed to do that. However, It’s also increasingly unrealistic to expect radiologists to manage their ever growing workloads (in size and complexity) without ever making a mistake or getting burned out.
Instead, we should come to understand that people working with AI technology can outperform either one working alone. I believe AI systems hold the key to identifying missed findings and improving the experience for healthcare providers and patients.
2) Tackling pain points early
Radiologists are first and foremost, physicians and scientists. Our north star has always been providing the most accurate, timely and safe insights and outcomes for our patients and our clinical colleagues. Our trust can only be maintained by embracing AI when the evidence shows it can make a real difference in our day-to-day work. For example, if AI can help radiologists tackle some of the routine, repetitive, ripe-for-error tasks – such as measuring the size of lymph nodes – it can help radiologists not only improve efficiency, but also the quality of care we provide patients. But widespread adoption will also require that solutions make good financial sense for radiologists and the healthcare system. Showing incremental impact across these areas will help smooth the pathway to adoption.
3) Integration into the radiologist’s workflow
Having successfully transitioned from film and dark room into the digital world, radiology is at another time of transition – advancing the field with advanced analytics and AI. But a significant hurdle remains: integrating these new technologies seamlessly into the workflow. There are existing applications of AI, such as natural language processing (NLP) for example, that will be important to transforming radiology workflows. The creation of centralized, uniform standards can also help speed this transition.
4) Standards for safety and effectiveness
Radiology has always been an innovator when it comes to applying technology in medicine, and being an innovator requires rigorous reviews and blazing new trails when it comes to standards. Through a series of initiatives, the U.S. Food and Drug Administration (FDA) continues to signal that it recognizes the potential of AI to improve healthcare, and that the regulatory process must evolve its approach to address the complexities of AI.
The FDA held a public workshop in early 20201, on the Evolving Role of Artificial Intelligence in Radiological Imaging. This meeting brought together radiology and technology leaders to discuss safety and effectiveness of AI applications, such as how to monitor autonomous AI. These government-driven conversations, and the many others with standards-based organizations across the medical imaging community, are critical to the development of the evidence-based decisions that will further enhance the speed and quality of care that AI promises.
5) Protection against potential ramifications of missed findings
Healthcare is still in its very early days of applying AI. There are a lot of difficult ethical questions, such as “If AI doesn’t detect an issue, or if a physician chooses to ignore an insight from AI, who is responsible for the missed finding?” or “Should we keep AI insights as part of the medical record, and if so, what happens if it includes early indications for a missed finding?”
Healthcare providers must address these and other ethical questions about how to use data for the good of patients. They must determine what to do with that data, now that there is so much more of it.
6) Recognition that AI can serve as an equalizer
Elite academic medical centers shouldn’t be the only places where providers and patients can benefit from AI insights. It’s time for the specialty to democratize innovation and enable informed decision making regardless of size or geography. It’s also critical that AI algorithms are developed with diverse data, free of race, gender or other biases. Otherwise, the quality, accuracy and reliability of the algorithms suffer, ultimately impacting confidence in AI as a solution. AI should be an equalizer, not a cause of further disparity.