Research
Two research projects: AI in high-stakes medical decision-making, and physics-based algorithms for 3D graphics. Both presented at undergraduate research conferences.
The Implementation of AI in the Field of Anesthesiology
Emmanuel Hernandez · Hinanui Swider
Irvine Valley College, Honors Program — Mentor: Prof. Kristen Skjonsby
The Problem
Getting anesthesia dosage wrong is catastrophic either way. Too little and a patient can wake up mid-surgery; too much and you risk respiratory failure. Anesthesiologists factor in weight, height, age, comorbidities, surgery type, and medication history all at once. We wanted to see how close a large language model could get, and what it would miss.
What We Did
We gave GPT-4 10 real patient profiles and told it to act as an anesthesiologist: output a recommended dosage and flag any risks. Then we compared every prediction against what the head anesthesiologist in Kauai, HI actually charted for the same patients.
Key Findings
- AI agreed on major risk categories but named them broadly; doctors flagged specific conditions like "aspiration risk" for individual patients
- ChatGPT recommended lower doses across the board, especially for Fentanyl and Propofol
- For complex patients with multiple comorbidities, AI was more cautious than clinical judgment warranted
- Best use case is first-pass monitoring and documentation, not replacing surgical decision-making
Research Summary
We looked at whether ChatGPT could do what an anesthesiologist does: take a patient profile and output a safe dosage recommendation with risk flags. Tested it against a real anesthesiologist in Kauai, HI across 10 patient scenarios.
Full Poster
Documents
Triangle Mesh Renormalization Using Physical Principles
Emmanuel Hernandez · Colin Minhquan Pham
Irvine Valley College, Department of Mathematics — Mentor: Lan Pham
The Problem
3D objects in simulations and games are made of triangles. When a shape changes over time, those triangles have to update too. The problem is that moving one interior vertex to fix a bad triangle shifts every triangle connected to it. Standard geometric algorithms handle this badly. We wanted to try a physics-based approach where the mesh finds a good configuration on its own.
What We Did
We modeled each interior vertex as a point mass connected to its neighbors by springs, with virtual charged particles along the boundary keeping points from escaping. The system evolves by applying Newton's 2nd law as a 2nd-order ODE, stepped with Euler's method. Code in VB.NET, visualization in MATLAB.
Key Findings
- Interior points converge toward equilateral configurations under spring and damping forces
- Electrostatic boundary repulsion kept 95%+ of interior points inside the region across all test geometries
- Tested on rectangles, triangles, L-shapes, and time-evolving shapes where the boundary changes mid-simulation
- Presented at the Bay Honors Symposium at UC Berkeley
Research Summary
Triangle Mesh Renormalization Using Physical Principles. Co-authored with Colin Minhquan Pham, mentored by Lan Pham, Dept. of Mathematics, Irvine Valley College.
Interested in collaborating?
Looking for research opportunities across CS, healthcare, and graphics.
Get in Touch