Despite historical advances in medicine, healthcare remains largely inaccessible around the world. For example, the proportion of individuals with Parkinson’s who are undiagnosed ranges from 12% in Rotterdam to 100% in rural parts of Latin America.1,2 In the U.S., only 9% of individuals with Parkinson’s disease see a movement disorders specialist; 40% do not see a neurologist at all. Those that do not have worse health outcomes.3 Emerging tools, namely artificial intelligence, could help close these gaps by extending the reach of clinicians and increasing access to care.
Artificial intelligence has already been incorporated as an integral component of our daily lives, in ways we often don’t notice.4 Google Maps uses neural networks to predict and adjust our estimated time of arrival while we drive. Social media platforms like Instagram and TikTok use algorithms to determine the content we see. Virtual assistant technologies like Amazon’s Alexa and Apple’s Siri have become household names, seamlessly completing tasks as requested. And AI will only become more ubiquitous: sweaters will determine heart rate, mattresses will identify sleep disorders, smartwatches will diagnose neurodegenerative diseases.
Medicine has long established the use of AI and demonstrated its promise for future advancements. As early as 2007, artificial intelligence has been used to interpret voice sounds to identify Parkinson’s with up to 99% accuracy, as demonstrated by the brilliant mathematician, Dr. Max Little.5 Many studies have shown how smartphones and wearables can be used to not only detect symptoms, but also provide more objective measures of neurodegenerative disease.6,7 Most recently, computer scientist Dr. Dina Katabi used AI-enabled radiowave sensors to track breathing signals at night, detecting Parkinson’s in some individuals up to six years before a diagnosis.8
Though AI shows great promise within medicine, there are some shortcomings. In addition to the growing pains that Dr. Abhimanyu Mahajan notes in his excellent piece, concerns with privacy, equity, and duty must be addressed first.9
With regards to privacy, the Genetic Information Nondiscrimination Act protects individuals from discrimination based on family history and genetic information. A similar Data Information Nondiscrimination Act is needed for AI to protect the privacy of health information often collected by for-profit parties. Additionally, technology should be viewed as a public good, available to all. AI has the potential to reduce the substantial inequities in healthcare. It can only do so if all individuals can access them. Finally, a major concern is that these technologies will be applied outside of the health care setting where duty to shareholders, rather than patients, rules supreme. The interests of our patients requires protection.9
Even with these concerns, artificial intelligence is poised to improve and play a crucial role in the diagnosis, treatment, and prevention of disease. Previous advances in technology within medicine have opened more doors and enabled advancements to improve care. For instance, MRI has completely changed the way multiple sclerosis was diagnosed and monitored, allowing for improved treatments and reduced morbidity. The same will remain true for artificial intelligence. AI will bridge critical gaps by increasing access to care on a global scale and reducing barriers for both patients and clinicians. More importantly, AI will augment the abilities of neurologists, allowing them to redirect their efforts to improve patient care and push the field in ways previously thought not possible. As such, artificial intelligence should be welcomed within the practice of medicine. In doing so, we can address structural inequities in healthcare, nurture medical advancement, and improve the wellbeing of our patients.