Feature article

Expanding Imaging AI to Resource-Strapped Areas

With AI at the helm, patients can benefit from a more streamlined experience, an advantage made all the more meaningful for those in underserved communities

Artificial intelligence has been shown to have lasting impacts on the way healthcare systems work together to improve patient care, especially from an imaging perspective. Thanks to powerful algorithmic learning, AI has made it easier than ever to pore through large volumes of scans to spot diagnostic patterns on the clinical side, and also improve workflows on the operational side.1

Each of these benefits—both clinical and operational—can enhance the patient experience before, during, and after every care interaction, even if patients may not ever realize that AI had a hand in doing so. In larger, more resource-rich metro areas with ample technologies and radiologist support, AI can make a considerable difference.

But it’s the patients from smaller, underserved areas—areas more strapped for staff, technologies, and time—that stand to gain the most from such technologies, and maybe should: after all, part of GE Healthcare’s AI principals reinforces a commitment to diversity. That diversity, gleaned in AI data sets that mirror the global population, is what helps remove potential biases like those that may favor urban or high-income populations. When all people are represented, all people can be helped—even when health systems are limited in human talent.2

That’s where AI can help step up to the plate, says Paul Edwards, Performance Intelligence and Analytics Marketing at GE Healthcare.

“Radiology resourcing as a whole, from a global context, is shrinking,” he said. “We have fewer radiology specialists coming into the fold overall, but the patient need is still there with workload and time pressure increasing for clinicians.”

From pneumothorax to prenatal care, and beyond

Edwards describes a potential for of an X-ray algorithm that can detect pneumothorax, or collapsed lung, within seconds.

“Imagine if you get priority notification for an image to be read,” he said. “It’s like having a priority pass where you wouldn’t have one before.”

Consider this from the patient’s perspective—especially on a global scale, adds Dr. Mathias Goyen, Chief Medical Officer, Europe of GE Healthcare. He gives an example of AI’s potential on prenatal care, an area of great clinical need throughout low-income populations.

“Say you’re pregnant in rural India with a lack of doctors,” he said. “You might have midwives performing ultrasound exams, and the built-in AI capabilities within that ultrasound could help detect complications or diseases before they evolve into life-threatening states.”

At that point, the provider can assess whether the woman needs further treatment in a hospital or if it’s something that her rural clinic can maintain. If the latter is the case, Dr. Goyen added, it may save the expense of having to travel hundreds of miles away to go to the nearest metropolitan hospital.

Such scenarios require those community clinics to have advanced technologies with built-in AI—but the portability of emerging equipment makes that more possible than ever.

But even if underserved areas lack hospitals with AI technologies, telemedicine can democratize AI on a broader, all-encompassing scale, he added. In that case, the same woman gets the same prenatal ultrasound, but the images can potentially be reviewed and plugged into the AI system from the big-city hospital on a remote basis.

“All of a sudden, AI opens up new opportunities for patients who can get scanned in more places, at more times, and by more people,” he said.

Advancements in non-pixel AI

All of these advancements are thanks to what the radiology sector calls “pixel AI,” or the clinical applications of imaging AI. But Edwards also points to advancements in “non-pixel AI”—meaning the workflows and operations that support care delivery, from the patient to practice. Though patients don’t see these technologies first-hand, the downstream impact yields an improved patient experience for underserved populations via reduced wait times and a more seamless care journey.

“Non-pixel AI has huge potential to provide more value to organizations and patients, because of the benefits to the workflow,” Edwards said. One such opportunity that GE Healthcare is looking to leverage in AI involves "no-show" prediction.

All told, no-shows have a marked impact on radiology organizations, with some experts pricing that opportunity cost around $1 million a year.3 Edwards suggests AI technology could help in a significant way.

“Essentially, you can leverage customer data, properly aggregated, within an imaging department framework and leverage AI to take into account a variety of factors using internally sourced data, but also externally sourced data like weather,” he said. “Through this method, we discover the probability that certain patients will arrive at their scheduled time or if they’ll be a no-show.”

Through those efforts, practices may be able to better anticipate and allocate resources to serve the patients who do show, and not the ones who don't. From the patient’s perspective, they'll have reduced wait times for exams, while not being conscious of the increased utilization that the customer sees. This benefit is made all the more meaningful for those in underserved communities who otherwise might not have access to them.

Rural or urban, underserved or not, AI is invisible if done well

Each of these applications—whether pixel or non-pixel, rural or urban, prenatal care or a collapsed lung—they all amount to AI working silently on the backend so that patients can benefit on the frontend, though they may not ever know it.

And that’s the point, says Dr. Goyen. Done well, AI should be invisible. With or without machines, it’s all just medicine, anyway.

“Nowadays, we talk about precision health, precision medicine, molecular therapies, you name it,” he said. “I hope that very soon, it’ll just be called medicine, because that’s what it is. AI will continue to have a growing role in healthcare, and in time, it’ll just become a normal part of the care continuum.”

References:

  1. How AI and Deep Learning are Revolutionizing Medical Imaging. GE Healthcare. https://insights.gehealthcare.com/medical-imaging/how-ai-and-deep-learning-are-revolutionizing-medical-imaging/. Accessed 10 May 2019.
  2. Ethics in Healthcare Aren't New, But Their Application Has Never Been More Important. GE Healthcare. http://newsroom.gehealthcare.com/ethics-healthcare-arent-new-application-important/. Accessed 10 May 2019.
  3. Appointment No-Shows Could Cost an Average Radiology Practice $1M a Year. Radiology Business. http://www.radiologybusiness.com/topics/care-delivery/annual-uncaptured-revenues-radiology-exams-could-equal-1m. Accessed 18 June 2019.