Artificial Intelligence and the Augmented Practice of Medicine
© 2023 HMP Global. All Rights Reserved.
Any views and opinions expressed are those of the author(s) and/or participants and do not necessarily reflect the views, policy, or position of Podiatry Today or HMP Global, their employees, and affiliates.
With the release of ChatGPT by OpenAI, the general public has begun to explore the implications and impact of a future with artificial intelligence (AI) at their fingertips. What once was science fiction is now coming into a very real focus, but there is a rising fear and skepticism of a coming world where human-centric work becomes obsolete.
In my observation, medicine is perhaps the highest pillar that technology would need to climb to advance beyond the need for human intervention. However, health care professionals should already know that lower-tier processing systems have already been part of the health care experience for decades. Although the term “AI” is thrown a lot as a new buzzworthy term applied to any system of data processing, in my experience, there are distinctions and levels of processing that one must understand to fully gauge the timeline of the impact to the medical profession.
Health care providers and their professional societies recommend treatment plans for certain diseases, which one can extrapolate as an algorithm. It is important to remember that AI and algorithms are related concepts, but they have distinct differences. An algorithm is a step-by-step set of instructions or rules to solve a specific problem or to perform a task. Algorithms have been used for centuries in various fields and are essentially predefined procedures or calculations. In health care, the more complex the disease, the more delicate and deliberate the algorithm will likely be. As the complexities of certain diseases and their treatments or prevention are better understood, algorithms simply help with the tasks of triage or care assignments. In other words, one can reduce an algorithms to a limited directional process that provides routing information, much like a map. However, these “maps” obviously have limitations, lacking the ability to extrapolate and create unique ideas.
Improvement upon simple algorithmic processes began with machine learning. In 1949, Donald Hebb described the communication process between neurons while touting theories of artificially created neural networks.1 This expanded in the 1970s and 1980s as computers began to increase the speed of their processing.2 For decades now, machine learning, big data, and neural network algorithmic processes have been touted by health care futurists as assessable predictive engines for prophylactic preparedness against disease outbreaks.3 However, as ChatGPT, and similar programs have prompted, diagnostics, increased productivity, and individualized treatment plans may be in the first wave of benefits for health care providers.4 In my observation, it is this promise, that AI may potentially enable machines to learn, reason, and make decisions similar to (if not better than) humans.
While algorithms are deterministic and follow a predefined set of rules, I find that AI algorithms can potentially adapt and improve their performance based on data and experience. AI algorithms can learn from patterns, make predictions, and generalize knowledge to new situations. They can handle complex and unstructured data, learn from large datasets, and discover insights that may not be apparent through traditional algorithms. Algorithms are specific sets of instructions to solve a problem, while AI involves the broader concept of machines mimicking human intelligence and learning from data to perform complex tasks.
However, this requires huge amounts of data, powerful, expensive computers to process this information, and time to trial and determine if the results are significantly better than previously available. For example, the average cost for 4 years of medical school in the US is nearly $250,000.5 Residency training can be 3–7 years with or without a few extra years for fellowship training. In total, these costs to train doctors to independently join the work force may be close to $1 million. The average cost for OpenAI to run ChatGPT version 4 is $700,000 a day.6 This does not include the costs of creation of all other generations of ChatGPT. Although many corporate companies tout tailored AI software solutions through a lease of an AI program, these leases can still cost about $500,000 a year.7 Although AI may exist in the health care space, it may not be as cost-productive as training new health care providers.
It's important to note that while AI in health care may offer numerous benefits, I believe it should complement the expertise and judgment of providers, rather than replace them. Ethical considerations, data privacy, and ongoing research are crucial in effectively harnessing the potential of AI in health care. Currently, there are numerous examples of health care providers utilizing AI in practice today.
In the field of radiology, computer-aided detection was first cited in 1992 to detect microcalcifications in mammograms.8 This groundbreaking study led radiologists and health care providers to imagine how AI could help reduce the workload burden, providing accurate diagnoses. AI-based systems such as workflow automation became adopted by most hospitals and helped clinicians with real-time decision support by analyzing patient data, medical records, and clinical guidelines. Automating operational tasks, such evaluating imaging quality, patient coordination, and improvement of disease reporting are now in several hospital systems with the assistance of AI.9 These systems have offered treatment recommendations, alert physicians to potential drug interactions or adverse events, and aid in identifying the most effective treatment options. In my experience, this has become vital, especially in light of post-COVID-19 staffing issues.
The increasing usage of virtual chatbots is another example of adaptive medicine with AI due to staffing issues post-COVID. These AI-powered virtual assistants and chatbots can provide basic medical information, answer patient queries, and schedule appointments. Also, as more and more patients turn toward the Internet for health care information, care seekers are becoming more comfortable with asking virtual chatbots for personalized health recommendations based on symptoms.10 Although tests comparing answers by virtual chatbots like ChatGPT versus actual physicians have shown mixed results, some patients preferred the chatbots due to higher perception of empathy in the answers.10
Wearable devices that monitor various health parameters, such as heart rate, blood pressure, and sleep patterns have also become a field where AI can improve health care. These devices can perform many functions including tracking changes, providing health insights, and alerting providers when intervention is required. The reason for heavy investment by companies in the development of such devices is that AI for individualized care requires lots and lots of data. Wearable devices that are worn hours throughout the day and night can track changes and monitor vital signs constantly, thus improving AI recommendations.11
These examples demonstrate how AI is already being used across different aspects of medicine to improve diagnosis, treatment, monitoring, and patient engagement. The adoption of AI technologies continues to evolve, with health care providers exploring new applications and expanding the integration of AI into their practices. More importantly, AI is still decades away from replacing actual human doctors, if ever, but it could be increasingly important for health care providers to use AI in their respective fields. Much like a revolutionary medical treatment, I feel that AI is becoming a critical tool to augment the practice of medicine, and thus, practicing physicians should not avoid engaging with this evolving technology.
Dr Sung practices in Arcadia, CA. He discloses that he is an early seed investor in Ahura AI.
References
1. Hebb DO. The Organization of Behavior: A Neuropsychological Theory. New York: Routledge, Taylor & Francis Group, 1949.
2. Foote KD. A brief history of machine learning. DATAVERSITY, December 3, 2021.
3. Mayer-Schönberger V, Cukier K. Big Data: A Revolution That Will Transform How We Live, Work And Think. 2013.
4. ChatGPT. The implications of on podiatry practice. Podiatry Management Online, 2023.
5. Hanson M. Average Cost of Medical School. EducationData.org, May 18, 2023.
6. Chowdhury H. ChatGPT cost a fortune to make with OpenAI’s losses growing to $540 million last year, report says. Yahoo! Finance, May 5, 2023.
7. Palokangas E. How much does AI cost? What to consider. Scribe, May 23, 2023.
8. Driver CN, Bowles BS, Bartholmai BJ, et al. Artificial intelligence in radiology: a call for thoughtful application. Clin Transl Sci. 2020; 13: 216–218
9. Cowen L. How artificial intelligence is driving changes in radiology. Inside Precision Medicine, February 14, 2023.
10. Staff, CBS Baltimore. Why are patients turning to artificial intelligence chatbots for medical advice? CBS News, May 11, 2023.
11. Rosen H. Council Post: how generative AI can improve personalized healthcare with wearable devices. Forbes, April 17, 2023.
Disclaimer: The views and opinions expressed are those of the author(s) and do not necessarily reflect the official policy or position of Podiatry Today or HMP Global, their employees and affiliates. Any content provided by our bloggers or authors are of their opinion and are not intended to malign any religion, ethnic group, club, association, organization, company, individual, anyone or anything.
