Summary: AI Revolution in Medicine, third of a series of four articles on Artificial Intelligence from The Harvard Gazette [i]. Published November 11, 2020. By Alvin Powell, Harvard Staff Writer.
Artificial Intelligence is “the theory and development of computer systems able to perform tasks that normally require human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages” [ii].
When some people think of AI, their minds immediately go to a dystopian form of the future. Other people see a utopian world where doctors from even the world’s poorest places can deliver world-class care more efficiently with little cost. AI is not perfection, but the truth seems to be closer to the latter than the former. Experts agree that AI will be a force multiplier, easing the workload while improving the outcomes. For example, most people can successfully manage four tasks at once. When a fifth task is added, everything falls apart. This is where AI can be a true human lifesaver. AI supports medical teams when the fifth or more thing is added to their workload. At this point, little details can be overlooked that a medical team may not notice. When reminded by an AI system, the medical team can pick up on the details. A study conducted in 2016 showed that AI systems alone could correctly identify 92% of cancer cells in pathology slides. Pathologists alone could correctly identify 96% of the same slides. However, the combination of AI and the pathologist correctly identified 99.5% of the slides [iii]. AI is not meant to be used alone, but as part of a team.
In the short term, the biggest roadblock to using AI successfully has been a lack of reliable, real-time data caused by delays in data collection and privacy concerns. Not all doctors or public health facilities have up-to-date technology that allows real-time data sharing. Some places are still using fax machines. The COVID-19 crises have highlighted these problems. An AI system being used in Wuhan, China detected an “unusual pneumonia” one full week before the World Health Organization. However, because of the lack of data sharing, researchers could not fully work to solve the problem more quickly. What researchers have learned from COVID-19 will be key to manage future pandemics better.
Longer-term issues with AI tend to be more social and cultural than technical. AI needs to be able to address cultural biases. AI is only as good as the people programming it. Having a more diverse programming workforce will be vital to the overall success. This is why AI needs to be tested under real-world conditions and re-evaluated regularly by the people who use the systems, no matter their job descriptions. These users need to have the final say. A medical worker in India’s rural area will know their patient better than an AI system created in Silicon Valley. The medical worker must have an override. If not, they may ignore using the AI system, missing other important information.
Considering there are issues with AI, overall, it is a win for the medical community, including patients. There are currently places in the world where it is more dangerous to go to the local medical facility than it is to be sick. With the assistance of AI, the opportunity for overall healthcare improvement is massive. It not only helps the doctors perform better, but also increases the chances for positive patient outcomes.
References:
[i] Powell, Alvin. “Risks and Benefits of an AI Revolution in Medicine.” Harvard Gazette, Harvard Gazette, 11 Nov. 2020, news.harvard.edu/gazette/story/2020/11/risks-and-benefits-of-an-ai-revolution-in-medicine/.
[ii] Lexico by Oxford, (n.d.). Artificial Intelligence, In lexico.com dictionary. Retrieved January 11, 2021, from https://www.lexico.com/definition/artificial_intelligence
[iii] Kritz, Jennifer. (2016, June 19). Artificial Intelligence Achieves Near-Human Performance in Diagnosing Breast Cancer. Beth Israel Deaconess Medical Center. https://www.bidmc.org/about-bidmc/news/artificial-intelligence-achieves-near-human-performance-in-diagnosing-breast-cancer
