Faculty Editor: Professor Jeffrey DaCosta
“Some people call this artificial intelligence, but the reality is this technology will enhance us. So instead of artificial intelligence, I think we’ll augment our intelligence.” -Ginni Rometty5
Artificial intelligence is a broad term that generally refers to computer systems and models that are designed to replicate human intelligence and abilities. The use of of artificial intelligence (AI) has increased over the last decade, yet many still oppose its use, primarily based on lack of knowledge of the technology, and the subsequent fear that AI will eventually replace people in many jobs. For example, in the medical field, there is a fear that AI machines will replace doctors, rendering physicians unemployed, and ultimately useless6. However, even as the use of AI in medicine increases, often the AI machines must work in conjunction with physicians, rather than working in place of them.
AI systems are increasingly being used in a diagnostic role within clinical settings. Engineers have been able to develop AI machines that analyze large amounts of previous patient records, primarily their symptoms and subsequent diagnoses, and then use that information to diagnose new patients. These machines often learn as they encounter new patient information, resulting in increased diagnostic accuracy. In essence, the machines mimic the way a physician would process relevant medical information and then come to a diagnosis.
At Guangzhou Women and Children’s Medical Center in China, researchers developed an AI-based system in order to test the accuracy of such machines in diagnosing common pediatric conditions3. Researchers used existing patient records to train the system. Then, to compare the accuracy of the AI system with real physicians, they inputted different patient records into the system, and allowed the system to come a diagnosis. Then, they compared the system’s diagnosis with the actual physician’s diagnosis. In total, over 1 million existing pediatric records were used to train and validate the accuracy of the system3.The system proved to have accuracy “comparable to experienced physicians,” especially in common conditions like acute respiratory infections, bronchitis, and tonsillitis.
Similarly, in Pittsburgh, PA, doctors used a machine learning algorithm, a form of AI, to create a model that calculates the probability of lung cancer based off of CT scan data2. About a quarter of low-dose CT scans indicate a positive result for lung cancer, but in reality fewer than four percent of those with a positive result actually have lung cancer. That is, 96% of people who are told that their scan is positive for lung cancer have actually received a false positive2. In order to test the accuracy of the model, the algorithm was given data from patients who had been confirmed to have benign or malignant nodules, and the diagnosis by the model was compared to the real diagnoses. The model was able to detect that 30% of the patients who tested positive for lung cancer did not have cancerous tumors. In other words, 30% of the people who in reality were given a false positive, would not have been given a false positive result if the model had been employed at the time of diagnosis. Remarkably, the model also did not miss a single case of cancer. Preventing false positives helps save patients from having unnecessary concern and anxiety, and also prevents them from having to undergo further testing, which in turn saves patients money.
AI models show promise for the medical field. Their diagnostic abilities tend to be the same as or even better than an experienced physician’s abilities. However, rather than replace physicians, these machines could be used to augment a physician’s abilities. Physicians are often needed to interpret a diagnosis, and to work with a patient in a way that currently, even the most intelligent AI systems cannot. Dr. Eric Topol, a prominent physician who recently released a book on the subject of the use AI in medicine, believes that AI will allow physicians to return to being able to spend more time with their patients rather than having to complete “mundane tasks” like typing medical information into patients’ files1. He believes that AI will allow doctors to be able to return to connecting with patients, improving the quality of care for patients, and may also help improve the mood and job quality for doctors themselves. AI used in conjunction with physicians may allow for a higher standard of care for patients, and may ease the demand of a physician’s job.
AI technology may the key to humanizing physicians again. It will not replace, but will rather augment the role of the physician. Patients may receive better care, both in the fact that AI diagnostic systems may reduce medical costs, anxiety, and in the fact that doctors will be able to return to more personalized, interactive care.
- Abril, D. (2019, April 2). How Artificial Intelligence Could Humanize Health Care. Retrieved from http://fortune.com/2019/04/02/artificial-intelligence-humanize-healthcare/
- University of Pittsburgh. (2019, March 12). Artificial intelligence cuts lung cancer screening false positives. ScienceDaily. Retrieved April 12, 2019 from www.sciencedaily.com/releases/2019/03/190312195050.htm
- Liang, H., Tsui, B. Y., Ni, H., Valentim, C. C., Baxter, S. L., Liu, G., . . . Xia, H. (2019, February 11). Evaluation and accurate diagnoses of pediatric diseases using artificial intelligence. Retrieved from https://www.nature.com/articles/s41591-018-0335-9
- London, A. J. (2019, February 21). Artificial Intelligence and Black‐Box Medical Decisions: Accuracy versus Explainability. Retrieved from https://onlinelibrary.wiley.com/doi/full/10.1002/hast.973
- Marr, B. (2017, July 25). 28 Best Quotes About Artificial Intelligence. Retrieved from https://www.forbes.com/sites/bernardmarr/2017/07/25/28-best-quotes-about-artificial-intelligence/#394f0164a6fc