Overview
In the summer of 2023, I joined the Philips Experience Design Team as a UX intern, creating the UI/UX for Philips' first AI-driven radiology report drafting system.
The goal of the project is to improve the accuracy and efficiency of the radiology diagnosis process, providing patient information and medical references within the radiology report drafting app.
Design Highlights
An efficient radiology report drafting tool with all medical references in one place
Context
All stakeholders in a radiology department
Junior radiologist are the primary user of a drafting tool
Dept. Head: customer of a drafting software
Jr. Radiologist: primary user, drafting the reports
Sr. Radiologist: secondary user, reviewing the reports
Report drafting process
Junior radiologists are responsible for analyzing data and images, then efficiently producing accurate reports that capture diagnostic decisions.
Problem Space
An inefficient report drafting experience
Multiple tools are required, making the job complex
Junior radiologists need to retrieve images from PACS and patient medical record from EMR, then using a separate report drafting tool to write the report.
Need to look at different apps when drafting
PACS: Image Archive
EMR: Patient Medical Record
Report Drafting Tool
The current drafting tools have problem with both UI & UX
Each hospital has its own drafting tool developed in collaboration with an outsourced agency which makes the tool usually outdated without maintenance.
To summarize:
Solution
Integrating all needed information in one place
Solution criteria
Patient information at a glance
Provide option to hide unnecessary information, reducing the cognitive load of reading
Organize information layout to improve efficiency of scanning material
Easy access to medical references
Provide a window for questions in any tab, assisting user to understand a specific term/ symptom
Support the answers with cited articles from data base, solidify the credibility of resource
Land on patient summary, switch tab to view the details
Tab 1: Patient Summary
An AI generated summary, briefing users about a patient's condition
Tab 2: Report History
A timetable of reports with highlights, providing decision making details
Tab 3: Knowledge Base
An AI chatbot, supporting user with easy-to-understand answers and cited sources
Tab 1: Patient Summary
High level summary as a primer
As users scanning through the text and gaining basic understanding of a patient's situation, they have the option to ask AI any related questions.
Design Choices
Tab 2: Report History
Detailed reports with highlights and access to the original files
The design uses highlights and filters to quickly sort relevant information and avoid bias, enhancing efficiency in identifying abnormalities.
Design Choices
Focus mode filtering
Design Process
Iteration
Feedback
Requires additional clicks to view highlighted information
Patient Summary+Report History overwhelming users
Filter does not accommodate the situation of finding several diseases
Final Design
Feedback
Landing with Patient Summary, providing a quick overview for user's information
Distinguishable highlighted report cards, suggesting areas to read
Filtering to focus on only related report, improving the efficiency of reading
Tab 3: Knowledge Base
Ask questions and refer to the medical references while writing
The scalable chat/ read window optimizes drafting and reading, enhancing legibility for users.
Design Choices
Design process
Iteration
User Test
“I found having a small window answering questions insecure”
“I would like to read the references without leaving the drafting app”
Final Design
User Test
Clear display of AI answer and external links
"The scalable window offers better legibility when I write"
Final Outcomes
Impact of the design
Before
No patient history as diagnostic supporting material
No domain references to support junior radiologists
Disrupted user flow & outdated 2000s UI
After
Patient history presented with a chronological table
Domain references presented as AI chatbot with access to articles
