Case Study: AI in Surgery
Designing a Context-Aware Assistant for Urogynecologic Procedures
Real-time guidance without distraction. Ethical, explainable AI in the operating room.
Overview
This project explored how AI could assist in surgical procedures—not by replacing expertise, but by enhancing it. I focused on a specific scenario: prolapse surgery within urogynecology. The result was a dual-mode AI interface that adapts to the needs of both surgical residents and attending physicians.
Role: UX Researcher & Designer
Tools Used: Figma, ChatGPT, Replit, Stanford AI coursework
Project Timeline: March – April 2025
Problem
Surgical training often lacks scalable, real-time mentorship. Surgeons—especially in complex or high-pressure procedures—need context-aware support that improves safety and learning without causing distraction or dependence.
How might an AI assistant provide helpful, real-time surgical guidance while respecting expertise, preserving autonomy, and maintaining ethical transparency?
Research
I conducted secondary research on surgical workflows, cognitive load, and AI use in clinical settings. I also gathered insights from informal interviews with surgical educators and residents.
Findings included:
Residents benefit from step-based, interactive feedback
Attendings prefer subtle, optional nudges over intrusive guidance
Transparency is essential—AI suggestions must be explainable and easy to dismiss
Design Approach
I designed a dual-mode AI assistant tailored to surgical roles and confidence levels.
Resident Mode
Layer-by-layer anatomical walkthrough
Voice-guided support at key decision points
Interactive questions like “What do you expect to see next?”
Pro Mode
Passive, sidebar-style guidance
Suggestions offered only at moments of known ambiguity or risk
Post-operative AI summary with annotated deviations and potential risk markers
Interactive Prototype
Using Replit, I created a simple decision-support flow simulation. It shows how the AI might adapt based on real-time surgical context and user role.
User toggles between Resident and Pro Modes
Prompts and suggestions vary by experience level
Sample post-op summaries generated with contextual feedback
Outcomes
Feedback from medical professionals highlighted the balance between automation and agency. They appreciated the restraint in design—especially in how suggestions were timed and explained.
Increased user confidence in Resident Mode scenarios
Positive response to post-op AI-generated insights
Emphasis on transparency and control resonated across interviews
Reflection
This project pushed me to consider design beyond screens—to think in rhythms, voices, and cognitive pacing. It challenged me to design for trust, especially in environments where every second counts.
It also sparked a deeper interest in how UX can shape ethical AI behavior in medicine—through tone, timing, and choice.
What’s Next
I’m continuing to explore AI + healthcare workflows—particularly how explainable AI and design ethics can be applied in robotic-assisted surgery, patient education, and post-op monitoring tools.