The Promise and Perils of Artificial Intelligence in Medicine

By: Kyle Lee and Renesh Gudipati

The quickly-developing field of Artificial Intelligence (AI) now permeates nearly every aspect of life. Even the sectors most reliant on the complexity of human cognition are not immune to its impact. One of these areas where AI’s influence is most curious is medicine. From enhancing diagnostics to optimizing drug research, AI has the potential to make significant positive impacts on the quality of healthcare [1]. However, like any new development, AI’s interplay with medicine does not come without a set of disadvantages. Exploring the pros and cons of AI in medicine can help future medical professionals to better understand the extent to which it can be integrated into their career.

As AI blossoms in the medical realm, there are many prospective benefits it supplies. AI is capable of manipulating large amounts of data, which makes it a valuable resource for research, providing real-time data for diagnoses and streamlining mundane tasks. In the clinical sense, it can also lead to decreased physician stress [2].

One case that exemplifies the application of AI to disease surveillance is the outbreak of COVID-19. It is evident how fast a virus can spread and how difficult it has proven to be to decipher its source. AI algorithms are currently used in early warning systems; they compile large amounts of outbreak reports and records from the web to formulate a database of potential hotspots to avoid and contain. In 2019, one of many such algorithms, HealthMap, was able to track vaping-induced pulmonary disease in the United States. Furthermore, AI algorithms have been brought to wearable devices like smart watches and rings which can aid in recognizing symptoms and exposure. AI’s usefulness in disease management will continue to develop and will be a huge asset when combating disease outbreaks in the future [3].

Many hospitals have implemented AI to streamline tasks such as appointment scheduling, reading insurance policies, and even bedside assistance. Automating these tasks has saved hospitals time, money, and stress because an AI interface can handle loads of information and make judgment calls. For example, AI algorithms use instance markers to identify mistaken insurance claims and automatically generate details which saves the hospital a lot of time spent reading the insurance claim and money for any claims that are missed. Specialized AI can also be trained to recognize certain images to assist in a prognosis for the patient: For example, the deep convolutional neural network (DCNN) is a model created by Google that classifies images as diabetic retinopathy and macular edema for adults with diabetes. Models like these continue to grow as research is conducted on AI and we can expect to see more AI assistance in patient prognosis and healthcare in the coming years [4].

While AI certainly has the means to improve the quality of healthcare in many regards, it is important to remain critical of the technology’s potential complications. Some challenges this new technology poses include data privacy and ethical considerations.

Central to the healthcare industry’s operations is the handling of sensitive personal information that can be compromised if AI is not securely implemented. Under the Health Insurance Portability and Accountability Act (HIPAA), patient data including medical history, identifiable details, and payment history is protected from exposure. These pieces of information are often fed to AI systems for them to operate efficiently. The problem is that when given such high volumes of data, there is a greater risk of leakage [5]. According to a report published by the HIPAA Journal analyzing healthcare data breaches in America, over 6.5 million records were breached as of October 2022 due to the inadequacy of AI security measures [6]. If AI is not able to protect the sensitive information it uses, it would threaten the very integrity of healthcare.

Other than privacy concerns, the use of AI also comes with a series of ethical dilemmas. For one, the results of AI rely heavily on the data they are built upon. Thus, if the original training data contains biases due to factors such as lack of diversity, the AI system may perpetuate the same biases [7]. Inequitable healthcare outcomes may follow suit. Another ethical qualm comes with transparency and liability. A key value of modern healthcare is the concept of informed consent — meaning patients have the right to fully understand their treatment and its implication before making a decision. With the complexity of AI, it may become harder to establish trust with and ensure maximum comprehension of patients. Furthermore, if AI generates errors or harm, it may become difficult to discern accountability using traditional models. [8]

To fully harness the benefits of AI in medicine while mitigating its drawbacks, collaboration between healthcare professionals, researchers, policymakers, and technology experts is crucial. By fostering responsible AI development and implementation, we can embrace the advancements AI offers and pave the way for a brighter, healthier future for all.

Kyle Lee is a third-year Neuroscience major at UCLA. Renesh Gudipati is a third-year Computational and Systems Biology major at UCLA. Kyle and Renesh are both THINQ 2023–2024 clinical fellows.

Sources:

1. Berkeley and Berlin. (2023, April 20). How ai could change computing, culture and the course of history. The Economist. https://www.economist.com/essay/2023/04/20/how-ai-could-change-computing-culture-and-the-course-of-history?utm_content=section_content&gclid=Cj0KCQjwib2mBhDWARIsAPZUn_nIfA_NDReO75RKFdCjfMinEd7Bj_6xnFfp8krZjN7vcUHrqaFHN58aApupEALw_wcB&gclsrc=aw.ds
2. College of Computing & Informatics. (n.d.). Pros & Cons of Artificial Intelligence in medicine. College of Computing & Informatics. https://drexel.edu/cci/stories/artificial-intelligence-in-medicine-pros-and-cons/
3. Brownstein, J. S., Radar, B., Astley, C. M., & Tian, H. (2023, April 27). Advances in artificial intelligence for infectious-disease surveillance. The New England Journal of Medecine. https://www.nejm.org/doi/full/10.1056/NEJMra2119215
4. Basu, K., Sinha, R., Ong, A., & Basu, T. (2020). Artificial Intelligence: How is it Changing Medical Sciences and its future?. Indian journal of dermatology. https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7640807/
5. Murdoch, B. (2021, September 15). Privacy and artificial intelligence: Challenges for protecting health information in a new era — BMC medical ethics. BioMed Central. https://bmcmedethics.biomedcentral.com/articles/10.1186/s12910-021-00687-3
6. Alder, S. (2023, January 22). HIPAA, healthcare data, and Artificial Intelligence. HIPAA Journal. https://www.hipaajournal.com/hipaa-healthcare-data-and-artificial-intelligence/
7. Srinivasan, R., Chander, A., & Srinivasan, Fujitsu America, R. M. (2021, August 1). Biases in AI systems. Communications of the ACM. https://dl.acm.org/doi/pdf/10.1145/3464903?casa_token=caYYJTBagNwAAAAA%3AvheA5CUwSg5BvbcpJFzPw1elfi5ECxySwjIwEzLME1VI3pcs9gbaJEZFUBDHehTA_tEo1jWnDQiSWg
8. M. Cascell, L. M. (n.d.). Risk management tools & resources. Artificial Intelligence and Informed Consent | MedPro Group. https://www.medpro.com/artificial-intelligence-informedconsent