Causality of Primary Biliary Cholangitis (PBC) on Systemic Lupus Erythematosus (SLE) and Development of an SLE Predictive Tool; Ensuring Equity in the Development of Health AI Algorithms

Artificial intelligence has captured the zeitgeist of the public in recent years due to the immense capability it holds to enrich our lives — or to ruin them. Of all the fields to which artificial intelligence (AI) could be applied, there is perhaps no realm with as much to gain or lose from AI as healthcare. AI could revolutionize healthcare, saving huge numbers of lives by improving diagnosis and developing individualized treatment plans for patients (Cerrato & Halamka, 2018; Hubbard et al., 2023). At the same time, the life-or-death stakes of healthcare require AI developers to proceed with exceptional caution; should healthcare AI algorithms fail to provide adequate care, the consequences could be dire.
I completed my capstone project in biomedical engineering in partnership with AMPEL Biosolutions LLC (“AMPEL”) to improve diagnosis of systemic lupus erythematosus (SLE). SLE is a chronic auto-immune disease that manifests very differently from one patient to the next. There are a wide array of symptoms that patients with SLE can experience, and SLE has been known to impact organs as diverse as the kidneys, joints, and even the brain (Kuhn et al., 2015). The severity of SLE also varies greatly; to some people with SLE, it is just an occasional discomfort, while others lose their lives to the disease. The fickle nature of SLE makes it very difficult for clinicians to diagnose. Late diagnosis and delayed treatment increase the risk of patient mortality and rate of serious complications like strokes (Sebastiani et al., 2016). As such, it is important that we seek ways to improve SLE diagnosis, and AI could be the answer. AI might be able to capture nuances in patient data that are too hard for human clinicians to see by themselves, allowing for improved diagnosis. The primary goal of my technical project was to develop an AI tool to aid clinicians in diagnosing patients with SLE — and, by doing so, to improve outcomes for SLE patients in turn.
My technical project hinged around the design of an AI algorithm applied to healthcare — that is, a health AI technology. I felt I needed to understand the ethical considerations behind health AI while conducting my technical project in order to ensure that my own health AI was built ethically, so I dedicated my STS thesis to this topic. A particular ethical concern surrounding the use of health AI is its possibility to exacerbate inequity if designed or used poorly. Healthcare has historically been restricted to those of certain races and socioeconomic statuses, and many experts fear that AI trained on historical healthcare data will reflect this bias against marginalized communities (Khan et al., 2023). However, despite equity being a major point of discussion in the field of health AI, there are few resources for AI developers as to how to ensure equity in health AI algorithms. To bridge this gap in the discussion, my STS thesis investigated the following question: how do we ensure that healthcare ML algorithms perform with equity, minimizing as opposed to exacerbating disparity in healthcare?

School of Engineering and Applied Science

Bachelor of Science in Biomedical Engineering

Technical Advisor: Timothy Allen, Kate Owen, Amrie Grammer, Peter Lipsky

STS Advisor: Joshua Earle

Technical Team Members: None