Client I NeuralMed
Role I Freelance Product Designer
Services I Research, UX/UI
NeuralMed is a Healthcare startup developing AI-based solutions to optimize the screening flow and diagnosis of patients in the hospital ecosystem. I teamed up with their group of data scientists to redesign Florence, a platform for medical image annotations.
THE ORIGINAL ONLINE STORE
THE UNSUNG HEROS
OF AI DEVELOPMENT
The task of annotating medical data for healthcare algorithms is a time-consuming and tedious work without any of the flare associated with sci-fi-like thinking.
THE CHALLENGE
NeuralMed’s team needed millions of verified results to train their machine. So, they decided to improve the interactive experience of Florence for the doctors who do this work.
Video of the old platform
DISCOVERY PHASE
Early on in the process, I conducted interviews with several radiologists and summarized
key findings that characterized who they are and their annotating experience.
THE USERS
Brazilian Radiologists who
specialize in XR and CT scans
Over 10 years of experience working
in hospitals and private clinics
Dedicating 8-10 hours a week to
annotate images for extra income
Motivated to label X amount of images each week to reach a set financial goal
THE EXPERIENCE
Repetitive Process - Skilled radiologists can label images within several seconds
Lack of Feedback - Annotators are not valued for being quick or accurate
Monotonous Work - Lists of datasets force the same type of annotation
Wrong Motivation - To speed the process doctors often add less labels
VISION AND
STRATEGY
Our goal was to create additional value for users by adding tools that encourage personal growth and involvement - such as progress bars, leaderboards and customized dataset opportunities.
IMPROVING FLORENCE
UX DESIGN PROCESS
OLD MENU SCREEN
-
Featured only Xray data ('Raio-X)
-
Allowed doctors to select between labeling images of children or adults ('Indantil / massa')
-
Featured earnings in a different window ('Checar atividade')
NEW MENU SCREEN
-
Features a selection of 'bundled' images including CT scans and NLP reports
-
Each set bundle has a set prize tag
-
Clear depiction of monthly earnings and relevant annotating stats
-
Illustrates the doctors 'ranking' in relation to the community
OLD LABELING PROCESS
-
Medical prognosis in text form with no 'free annotation' labels (like 'other')
-
Label annotation tool unclear (frames box tool, drawing shape etc)
-
Menu of view options also in textual format
NEW LABELING PROCESS
-
Medical prognosis menu presented clearly with relevant iconography
-
Progress bar showing completion rate
-
View menu presented using icons
-
Set prize and earnings featured on screen at all times
POSITIVE FEEDBACK
-
Upon set completion doctors receive feedback with new account balance
PROTOTYPING & VALIDATION
USABILITY TESTING
We built a lean version of the platform and tested our ideas with several doctors to evaluate their response and collect feedback.
Usability Testing with Dr. Alexandre Penteado
Usability Testing with Dr. Alexandre Bialowas
THINGS WE FIXED
FEEDBACK IMPLEMENTATION
LEADERBOARD
FEATURE
During the testing we discovered that the financial leaderboard is making doctors feel uneasy, since speaking of 'earnings' is considered taboo in Brazil. So, we decided to change the ranking system to feature only XP points.
CALL TO ACTION
In the main menu, some doctors said that our original 'PLAY NOW FOR' button feels as though it is something they need to pay for.
LABELING OPTIONS
Following the feedback of doctors we also decided to give doctors the option to add labels which are not predefined ('Other').