2025 ASEE Annual Conference & Exposition

Many engineering programs have responded to the rise in interest and application of AI/ML by finding ways to introduce them in undergraduate curricula. This particular area of curricular change arose as one of the key concerns for faculty at the 2024 Biomedical Engineering Education Summit. The question was not if AI/ML should be included, the question was how. One pushback to such curricular changes is well established - concerns about space and about fitting all the content in. Our department has engaged in a curriculum-wide effort to add data skills - including but not limited to the use of AI/ML tools - in each required undergraduate course.

In this paper we share the development, implementation, and evaluation of one such curricular intervention. The intervention is a two course session (4 hour) activity for an introductory systems physiology course for biomedical engineering undergraduates. The guiding vision for this activity was to prioritize integration of the AI/ML content into the existing course in ways that (1) removed no content, (2) had students use real data and AI/ML tools, and (3) increased learning on the underlying course content. The activity relies on public data and an existing open science competition - the 2017 PhysioNet Computing in Cardiology Challenge. Our overarching approach is guided by theories of inquiry based and interactive learning that are well accepted in engineering education research.

In the activity, students develop and extend their knowledge of cardiac electrical signals and behavior, then they work in pairs to manually evaluate and classify ECG readings as if they were clinicians. They then reflect on what identifiers, data, and heuristics they are using to classify ECG readings, linking those decisions to their knowledge about the cardiac system. Next, they repeat the classification process - but using prewritten Matlab code and training neural networks multiple times on versions of the same ECG data. In doing so, students encounter and grapple with the challenges of visual analysis vs. analysis grounded in quantification of ECG signals, signals processing, and how what they see in the data can and cannot be converted into machine interpretable data using mathematics. They end by discussing the accuracy and validity of different techniques for diagnosing heart attacks via ECGs. We will evaluate the impact of the intervention using qualitative feedback from students and instructors as well as a pilot instrument that we are developing in parallel to this intervention as part of the larger, curriculum-wide effort.