In January of 2020, DeepMind and Google Health published research in Nature testing the effectiveness of their artificial intelligence (AI) technology in diagnosing breast cancer from mammogram images (1). Here’s the catch: the AI system outperformed human radiologists.
Although breast cancer remains the most common type of cancer and second leading cause of death for women worldwide, about 20% of screenings report false negatives, failing to diagnose the cancer when it is present in mammogram images. Similarly, about 50% of women who get annual mammograms receive at least one false positive every ten years (2). In DeepMind and Google Health’s study, however, AI technology was able to substantially decrease both false negatives and false positives across US and UK patient samples. Needless to say, the results are promising.
This is not the first instance of AI technology being applied to medicine. In 1986, an AI decision support system called DXplain was developed by the University of Massachusetts to fill in the knowledge gaps of medical textbooks and help medical students formulate diagnoses based on complex clinical findings (3). Similarly, the University of Sydney developed an AI-based cognitive behavior therapy intervention meant to help patients treat their social anxiety (4). That’s only the tip of the iceberg; AI has already begun to proliferate across diverse facets of healthcare including drug development, health plan analysis, digital consultation, and even surgery (5).
From an economic standpoint, healthcare investors are also pouring more money into AI technologies. From $2.7 billion across 264 deals in 2018, investments in healthcare AI startups surged to $4 billion across 367 deals in just the following year (6). As demand for high quality and up-to-date data only appears to be increasing over time, this upwards trend is likely to continue for many years.
First coined in 1956, AI describes the use of computers and technology to simulate intelligent behavior similar to that of a human being (5). The process of creating successful AI systems involves developing predictive, pattern-recognizing algorithms that are then trained on vast swathes of data. Generally, the tasks that AI handles are often arduous and require some level of human-like intelligence, such as analyzing images, recognizing speech, and making decisions (7). While the 1980s and 1990s saw a boom in interest in AI (5), healthcare lagged behind other industries in accepting and applying this new technology (8). Change, however, is around the corner as electronic health records become more widespread and medical data becomes more accessible. In a field where accurate and efficient predictions can—quite literally—save lives, the appeal of AI cannot be understated.
Nevertheless, full integration of AI in healthcare still has a long way to go. Healthcare is full of complexity and nuances, and current AI research standards are lacking (8). Given the resulting inconsistency of studies, as well as the lack of regulations around implementing findings, it is somewhat unsurprising how physicians and patients alike are sometimes reluctant to trust these new technologies (9, 10). After all, these algorithms lack the fundamental empathy and instinct that often defines the physician’s role.
That said, it is for this very reason that AI is necessary. As it is, physicians spend far too much time doing paperwork and updating the electronic health record system rather than actually treating patients. In 2016, a study published in the Annals of Internal Medicine discovered that for every hour that doctors interacted with patients, they spent nearly two additional hours on paperwork (11). Worse yet, that burden has only increased over time.
In order to remedy this onerous system, we need AI technologies. Whether these new tools make disease diagnoses more accurate or streamline the slow documentation process, they will simultaneously save time, money, and lives when applied in tandem with physicians and other healthcare personnel. Through this seamless integration, intelligent machines will not substitute in for the interpersonal role of doctors. Rather, they will lessen physicians’ other burdens, giving them more time for the very humanistic patient care that only they can provide.
With this new healthcare landscape inevitably upon the horizon, it is pointless to resist or question the feasibility of AI's emerging role in medicine. Instead, it’s finally time for all of us to begin engaging in a new conversation: what can we—the patients, physicians, policymakers, and innovators of tomorrow—do to prepare ourselves for this future?
- McKinney, S. M., Sieniek, M., Godbole, V., Godwin, J., Antropova, N., Ashrafian, H., Back, T., Chesus, M., Corrado, G. C., Darzi, A., Etemadi, M., Garcia-Vicente, F., Gilbert, F. J., Halling-Brown, M., Hassabis, D., Jansen, S., Karthikesalingam, A., Kelly, C. J., King, D., … Shetty, S. (2020). International evaluation of an AI system for breast cancer screening. Nature, 577(7788), 89–94. https://doi.org/10.1038/s41586-019-1799-6
- Google’s AI breast cancer screening tool is learning to generalize across countries. (2020, January 3). MIT Technology Review. Retrieved July 1, 2020, from https://www.technologyreview.com/2020/01/03/238154/googles-ai-breast-cancer-screening-tool-is-learning-to-generalize-across-countries/
- London, S. (1998). DXplain: A Web-based diagnostic decision support system for medical students. Medical Reference Services Quarterly, 17(2), 17–28. https://doi.org/10.1300/J115v17n02_02
- McCall, H. C., Richardson, C. G., Helgadottir, F. D., & Chen, F. S. (2018). Evaluating a Web-Based Social Anxiety Intervention Among University Students: Randomized Controlled Trial. Journal of Medical Internet Research, 20(3). https://doi.org/10.2196/jmir.8630
- Amisha, Malik, P., Pathania, M., & Rathaur, V. K. (2019). Overview of artificial intelligence in medicine. Journal of Family Medicine and Primary Care, 8(7), 2328–2331. https://doi.org/10.4103/jfmpc.jfmpc_440_19
- Investors poured $4B into healthcare AI startups in 2019. (2020, January 22). FierceHealthcare. Retrieved July 1, 2020, from https://www.fiercehealthcare.com/tech/investors-poured-4b-into-healthcare-ai-startups-2019
- Artificial Intelligence in Medicine: Applications, implications, and limitations. (2019, June 19). Science in the News. http://sitn.hms.harvard.edu/flash/2019/artificial-intelligence-in-medicine-applications-implications-and-limitations/
- Does AI Have a Place in Medicine? (2019, November 11). Scientific American Blog Network. Retrieved July 1, 2020, from https://blogs.scientificamerican.com/observations/does-ai-have-a-place-in-medicine/
- Briganti, G., & Le Moine, O. (2020). Artificial Intelligence in Medicine: Today and Tomorrow. Frontiers in Medicine, 7. https://doi.org/10.3389/fmed.2020.00027
- Longoni, C., & Morewedge, C. K. (2019, October 30). AI Can Outperform Doctors. So Why Don’t Patients Trust It? Harvard Business Review. https://hbr.org/2019/10/ai-can-outperform-doctors-so-why-dont-patients-trust-it
- Lee, B. Y. (2016, September 7). Doctors Wasting Over Two-Thirds Of Their Time Doing Paperwork. Forbes. https://www.forbes.com/sites/brucelee/2016/09/07/doctors-wasting-over-two-thirds-of-their-time-doing-paperwork/