AI technology and algorithms have great potential in radiology, where image interpretation and workflows are often time-sensitive and labor-intensive, and leave little room for error.
During an August Becker's Hospital Review webinar sponsored by Change Healthcare, Sonia Gupta, MD, Chief Medical Officer of Change Healthcare, and Evan Kaminer, MD, Chairman of Radiology at Montefiore Nyack (N.Y.) Hospital and CEO of Hudson Valley Radiology Associates in New City, N.Y., discussed AI adoption in imaging, how AI helps care teams and what to look forward to in partnerships.
Three key takeaways were:
1. Use of AI in radiology practices is growing. This trend can be observed across two dimensions: image interpretation and workflow optimization. Within image interpretation, common use cases include identifying stroke, intracranial hemorrhage and acute pulmonary embolism. Within workflow optimization, typical applications are managing incidental findings, assistance with radiology reporting, decreasing scan time, improving image quality and accelerating administrative tasks.
Dr. Kaminer said that at Montefiore Nyack Hospital, image interpretation AI algorithms integrated into a prioritization-based worklist have reduced ED turnaround time for critical findings by 17 percent or more. Hudson Valley Radiology Associates uses AI algorithms applied to screening mammography to help detect breast cancers up to two years earlier than a radiologist alone.
2. AI can also augment radiologists by improving care coordination and reducing burnout. The care coordination space is another area where AI in radiology is having an impact. Because AI algorithms that identify positive cases can simultaneously trigger notifications to the patient's entire care team, they reduce delays in activating the necessary protocols.
Further, while radiology AI tools do not replace radiologists, they provide much needed assurance associated with negative findings — essentially functioning as a second set of eyes and confirming that a negative reading is not a false negative. "It's a big stress reliever to know that there was an algorithm looking over your shoulder to make sure you didn't make an error," Dr. Kaminer said, adding that having an AI tool instead of a second radiologist for confirmation purposes is more affordable.
3. When considering a partnership with an AI vendor, organizations can pursue different avenues to choose an optimal fit. One approach is to attend radiology shows and conferences, where AI vendors display and demo relevant products. Another is to research offerings at so-called AI marketplaces, which resemble app stores and can reduce implementation costs. Reconnecting with colleagues from residency, fellowship or those in other practices and inquiring what AI technologies they are using is yet another option, as is reaching out to AI vendors directly.
Still, it is critical to keep in mind why a partnership is sought in the first place. "The first step is to identify the issues you want to solve in your practice — and not just get AI for the sake of getting AI," Dr. Gupta advised. "Once you identify the specific problem you want to solve, the next step is to find resources to support [the AI technology-finding] journey through your practice.
As in many other medical specialties, AI is making meaningful advances in radiology. Whatever type of modality or partnership organizations decide to implement, it is paramount to keep radiologists' needs and workflow preferences in mind. "You want to make sure that whatever solution you choose, it has a minimal impact on the radiologist's efficiency; otherwise, you're not going to get a big buy-in," Dr. Kaminer said.
To register for upcoming webinars, click here.