Azure AI Heathcare :
Medical Imaging Search Experience
Project Overview
Designing a Medical Imaging Search Experience for Radiologists
Radiologists often turn to general-purpose search engines to find similar X-rays, MRIs, or CT scans—an approach that's slow, unreliable, and not tailored to clinical needs. At Microsoft, I led the end-to-end design of an AI-powered medical imaging search experience within Azure AI Vision Studio, built specifically to support confident, fast, and informed diagnosis.
This experience allows users to upload a medical image, retrieve visually similar cases using computer vision, compare results side-by-side, and read structured clinical reports—all within a single streamlined interface.
I collaborated closely with PMs, data and applied scientists, radiologists + clinicians from Harvard Medical School and Mass General Brigham to iterate on the design through user feedback and iterative testing, prioritizing accuracy, usability, and trust.
Role
Lead UX Designer & Researcher
Tools
Figma
Duration
April - July 2023
Problem Statement
Radiologists often face time-sensitive situations where they need to reference similar medical scans—yet existing tools and workflows offer little support. Many resort to general search engines, which are not optimized for clinical use, lack visual accuracy, and fail to provide relevant diagnostic context.These inefficiencies can slow down decision-making, increase cognitive load, and lead to diagnostic uncertainty—especially in emergency or high-pressure environments.
How might we harness AI to deliver relevant medical image matches in a way that feels intuitive, trustworthy, and clinically meaningful?
User and User research
Research method - Focus group interviews with 10-15 radiologists + clinicians at Harvard Medical School, Mass General Brigham, and Nuance after each iteration to get quick feedback and rapidly iterate.
Since this was the first time we were building a search experience specifically for radiologists and clinicians and there was shortage of time, I started by conducting group interview sessions to understand their current process and flow. The cliniciuans ands radiologists were very excited about the future of AI + Radiology!
I kept very open and constant messaging streams with product leadership, my feature crew, and my manager to ensure we were all on the same page at the same time. I ensured that everyone knew what and exactly when things were happening. Clear communication, transparency, and visibility is absolutely essential when it comes to high-visibility and very quick design work.
🎨 Design concept development
After synthesizing insights from our focus group interviews, it became clear that speed, clarity, and trust were critical to this experience. Radiologists wanted to feel in control—able to quickly locate, evaluate, and compare images with minimal friction, especially during high-pressure diagnostic scenarios.
To translate these needs into design, I began with low-fidelity sketches sessions that mapped the ideal flow—from image upload to decision-making.
I aligned early concepts with familiar interaction patterns to reduce cognitive load while introducing new features like region-based selection and side-by-side image comparisons. Given the time-sensitive nature of the project, I adopted a lean, iterative design approach. Each iteration was shared in weekly reviews with cross-functional partners—data and applied scientists, PMs, and engineers—to validate feasibility while keeping usability at the forefront.
Feedback loops with clinicians were also embedded after each round, ensuring that every version was grounded in real workflows.
Iterations for refining the designs
Here’s a look at the design iterations I created before we arrived at the final experience—well, final enough at the time 😉. Each iteration included a fully clickable prototype that covered the entire user flow for radiologists and clinicians, allowing us to test, learn, and evolve the experience based on real feedback.
And finally, lets put it altogether for final iteration.. ✨
User Feedback and Impact 🎯
Reflection
- This project deepened my desire to observe users in context—seeing how radiologists actually work day-to-day could unlock powerful insights for future iterations.
- Reinforced the importance of rapid prototyping and continuous feedback loops, especially when designing for high-impact, time-sensitive use cases.
- Realized early on that dark mode was essential for radiologists working in dim environments—this influenced visual hierarchy and contrast decisions from the start.
- Explored opportunities to merge adjacent tools into a more unified search experience, reducing tool-switching and improving workflow continuity.
- Began thinking beyond image summarization—what other AI-driven insights could be valuable for radiologists?