Artificial Intelligence has already existed in healthcare for decades, dating back to the 1950’s when researchers explored the value of computers in diagnosing medical conditions. Now with the improved capabilities of ML models, AI now has the power to revolutionize the healthcare industry.
In 2021 alone, investment into AI climbed up to 108% and nearly 20% of the $66.8 billion allocated in global funding for startups was dedicated to healthcare AI. There are a number of areas in healthcare that show promise from AI capabilities, including work automation, forecasting, and acceleration of breakthroughs in “inhumanely large problem spaces”.
One of the most promising potentials of AI is accelerating diagnostic rates and enhancing precision by transforming medical devices such as: AI-powered heart monitors, stethoscopes, body scanners, and blood sugar monitors.
When tested against real-world case records, Microsoft’s AI Diagnostic Orchestrator (MAI-DxO) correctly diagnosed up to 85% of New England Journal of Medicine (NEJM) cases, which is more than four times higher than a group of experienced physicians. In a case where seventeen separate doctors mis-diagnosed a child’s recurring back pain, ChatGPT suggested a condition no other doctor considered when given the child’s medical information. This condition was then later confirmed by a neurosurgeon, which guided them to correct this medical anomaly in the child.
With the ability to streamline mundane tasks, from administrative work to documentation, AI can effectively reduce the workload of clinicians and alleviate their burnout. The rise of digital scribes would allow for “ambient documentation” to be possible, recreating the doctor-patient relationship that opens the pathway to stronger connections by freeing them note-taking behind a computer. By reducing the burden on healthcare providers, this energy and attention can be re-directed to delivering more personalized patient care.
Despite AI’s promise in improved outcomes, there remains barriers to full implementation in healthcare settings. Leo Anthony Celi, a clinical researcher at Massachusetts Institute of Technology (MIT) cautions on the quality of AI outcomes:
“Simply adding AI-powered devices to an already flawed system won’t help. We must also address the structural inequities that exist in healthcare.”
The lack of reliable data in healthcare, due to fragmentations and inconsistencies in medical records, makes it generally difficult for AI algorithms to accurately identify or diagnose medical conditions. For example, in attempts to detect breast cancer from mammographies, recent reviews from 2021 found that 94% of AI systems were “less accurate than a single radiologist, and all were less accurate than consensus of two or more radiologists.”