Some of the most exciting new applications of artificial intelligence (AI) technology are in the healthcare industry, where innovations could save patients thousands of dollars – and potentially their lives as well.
Just days ago, Microsoft Corp. announced that it is partnering with several medical institutions to expand AI use “in the medical imaging space.” If successful, doctor offices and hospitals will be able to use “generative AI” to speed up medical image generation while simultaneously assisting doctors and technicians in early recognition of diseases like cancer.
According to Microsoft, “health systems spend an estimated $65 billion every year on imaging.” Approximately 80 percent of all hospital visits also include at least one imaging exam. Generating and analyzing a medical image can take hours, if not days, requiring tremendous resources in time and manpower from healthcare providers. With many medical institutions facing physician burnout and staffing shortages, Microsoft’s developments could be just the answer hospitals and patients are looking for.
The partnership is the latest innovation from the adoption of artificial intelligence into the healthcare sector. Over the last three years, AI platforms have exploded onto the national stage. Starting with the launch of ChatGPT on November 30, 2022, for the first time ever, everyday consumers could access and utilize the benefits of AI technologies. Initially, ChatGPT was limited to text conversations. Yet, shortly after its launch, AI platforms were able to generate images and videos, and are one day expected to be capable of producing full length feature films.
In that same period, data scientists began searching for new areas to apply the revolutionary technology. In July, a predictive AI program helped categorize Parkinson’s disease into three distinct types. Historically, the degenerative disease was thought to be one whose progression varied from patient to patient. The AI program, however, was able to break down Parkinson’s disease into three separate types based on how quickly symptoms progressed. In delineating the progression, the program was able to identify variables that led doctors to provide treatments that address the root causes of each subtype. This could make treatments for Parkinson’s much more targeted and effective.
Online retail giant Amazon also recently launched a partnership with GE HealthCare to help physicians improve diagnoses using AI. GE executives said that Amazon’s cloud computing services and generative AI models will help doctors more quickly and accurately diagnose ailments.
Another AI breakthrough from researchers at Emory University and the Cleveland clinic is helping in early detection of age-related macular degeneration, one of the leading causes of blindness in the United States. AI programs are able to read eye scans to detect the beginning stages of inflammation that doctors may miss.
Researchers at the National Institutes of Health (NIH) have also developed AI tools that could more precisely match cancer patients with the most effective drug regimens.
Artificial intelligence has even proved useful on the more mundane side of healthcare. Last month, the company Maverick Medical AI developed a program to manage patient insurance claims more effectively. The company notes, “As healthcare providers face potential staffing shortages and seek to reduce billing errors, eliminate backlogs, and minimize claim denials, the interest and adoption of our advanced real-time solution has surged significantly.” Their program functioned with 93 percent accuracy and could significantly reduce costly layers of bureaucracy that often leave patients frustrated and unable to receive the care they need.
However, medical artificial intelligence is still in its nascent stage and does contain potential drawbacks. For instance, the NIH released a study in which they gave an AI model more than 200 diagnosis questions and asked it to justify each answer. It did the same with a group of physicians. The AI made more correct diagnoses than the doctors, but struggled to explain why each answer was right.
In another vexing example, a doctor and the AI model were given a picture of an arm with two lesions. The physician easily deduced that the same disease caused both lesions, but the AI model didn’t. The study concluded, “Because the lesions were presented at different angles—causing the illusion of different colors and shapes—the AI model failed to recognize that both lesions could be related to the same diagnosis.” The National Library of Medicine (NLM) Acting Director Stephen Sherry stated that AI healthcare integration “holds great promise as a tool to help medical professionals diagnose patients faster…However, as this study shows, AI is not advanced enough yet to replace human experience.”
Nonetheless, AI could provide serious relief to overworked hospitals and medical centers and aid in the diagnosis and treatment of diseases. While the healthcare sector will always need the care and compassion that only other human beings can provide, new technologies offer the potential for genuinely transformative breakthroughs in care in the near future.
Andrew Shirley is a veteran speechwriter and AMAC Newsline columnist. His commentary can be found on X at @AA_Shirley.
