Designing a cutting-edge ultrasound solution that detects blood clots within minutes.

This is a brief overview of the project (5 min read)

PROJECT TYPE

Android app

TIMELINE

Dec 2022 – Jan 2023

ROLE

Founding product designer

COMPANY

ThinkSono

context

Transforming how blood clots are detected

AutoDVT is an ultrasound solution that uses artificial intelligence to guide non-ultrasound medical professionals, such as nurses and GPs, in performing a deep vein thrombosis (DVT) ultrasound exam to detect blood clots. The tech dramatically reduces the cost and time to diagnose patients by increasing the number of medical staff who can perform the exam, not only radiologists.

Designing a beta release for clinical trials

As the founding designer at ThinkSono, my role was to establish the user experience for AutoDVT from the ground up. The goal was to release a beta to kick off clinical trials, gather clinical data,  and go through the regulatory process to receive FDA clearance.

My main focus areas were:
‍

1

Establish a streamlined UX process to meet an ambitious deadline to roll out clinical trials.

2

Lead the UX design process end-to-end and other design fronts outside UX design, such as marketing and
video projects.

3

Collaborate with founders and engineering to define product vision, strategy, and roadmap.


    ThinkSono, a MedTech startup that makes smart ultrasound software, began with a focus on funding, training, and perfecting their machine learning algorithm, the company's secret sauce. Then, with enough traction and success, ThinkSono was ready to move into product development. I was tasked to design the company's first product, AutoDVT.

    impact

    This project was successfully designed and developed on time despite an ambitious deadline. We achieved a significant milestone by completing the core flows to release a beta version to collect clinical data for medical device approval and validating the product as safe, reliable and fast.

    30x
    faster diagnosis

    Simplifying the number of steps and reducing the cost to get a DVT diagnosis.

    AUTODVT pathway
    Current pathway
    ≈ 5 min

    Scans done by any doctor or nurse

    Blood test not required

    Low cost (≈$70) per scan

    ≈ 6 - 24 hrs

    Scans only done by a specialist

    Blood test required

    High cost (≈$200) per scan

    The problem

    Inefficiencies with the existing diagnostic pathway

    Compression ultrasound has proven to be an accurate and non-invasive modality for recognizing DVT. During the exam, a radiologist will apply pressure to the veins in the affected area using an ultrasound scanner. If the vein cannot compress, it indicates a blood clot.
    ‍
    However, a worldwide shortage of radiologists limits healthcare systems from diagnosing and treating DVT efficiently. As a result, patients often wait hours for test results due to the limited capacity of radiologists, making the treatment expensive, time-consuming, and only sometimes possible.

    The solution

    Empowering non-experts with AI guidance and seamless UX

    Today, exams are done exclusively by radiologists or medical staff with extensive training limiting the number of individuals who can perform the exam. AutoDVT's AI guidance empowers non-experts by highlighting the veins of interest and providing step-by-step instructions to capture the correct data in the proper sequence, increasing the supply of staff who can perform the exam.

    Here is how it works
    Find
    Record
    Review
    DiAgnose

    Step 1

    Finding the veins

    With AI-powered written instructions, chimes, overlays, and other feedback, the app guides the user to correctly position and move the probe to find the vessels of interest.

    Step 2

    Recording the compression

    After finding the vessels, a 10-second recording of the vessels' compression is captured to be sent to the radiologist for a final diagnosis.

    If the veins compress fully, it indicates no DVT.

    Step 3

    Reviewing the scan recording

    After recording the compression, the user reviews the clip and records whether the veins are compressed for a provisional assessment.

    Step 4

    Radiologist's diagnosis

    The provisional diagnosis is sent to the cloud dashboard for review, where a radiologist can log in remotely and make a final diagnosis.

    UX FLOwchart

    Validating the flow and concept early

    To confirm the design concept and high-level user flow efficiently without spending too much time, I created a UX flowchart. After several revisions, the engineering team, the CTO, and I agreed on the screens and features that the beta release would include. The UX flowchart was also effective in communicating and confirming my understanding of requirements and flow with stakeholders.

    Wireframes

    Designing the core experience first

    With a solid understanding and team consensus on project scope and flow, I explored different scanning screen concepts before mocking the rest of the experience. I focused on the scanning screen because it is the app's core experience, and the rest of the content and navigation could be designed around it.

    v1
    v2
    v3
    v4 chosen solution
    Mid-fidelity scanning screen exploration
    Refining the core screens

    After exploring scanning screen concepts, it became evident that filling all the functionality we contemplated into a single screen was challenging and likely to decrease the learnability of the app. Therefore, I split the content into two screens to make the tasks easier to accomplish and scanning more straightforward.

    Final exam PROT screen
    Final scan screen

    Design

    Usability issues and improvements

    This is a short list of usability issues found during user testing and how I fixed them.

    1

    Confusing CTA label

    During user testing, participants found the 'star compression' button confusing because the actual vein compression happens outside the app by applying pressure using the ultrasound probe. The action button triggers a ten-second recording of the ultrasound image, so 'Record compression' was more in tune with the action.

    Before scanning screen
    After scanning screen with new CTA label

    2

    Inability to cancel a recording right away

    After starting a compression recording, the app didn't offer a way to cancel it, forcing the user to wait and advance to the end of the flow before a retry option became available. Through user testing with novice users still getting familiar with the scan protocol, we noticed many of them showed signs of frustration.
    ‍
    I used the same record compression button that switched to a cancel action right after a compression recording had started.

    Before scanning screen
    After scanning screen with Cancel recording option

    3

    Ultrasound gel and latex gloves made interacting with the phone difficult.

    During nurse training sessions, we learned latex gloves and ultrasound gel made tapping on interactive elements difficult, so we harnessed the probe's hardware buttons to make it easier for users to interact with the app. First-time users see an educational modal introducing the feature.

    Modal with educational message
    How we used it

    Scanner hardware buttons made it easier and faster for users to interact with the app. For example, it became highly convenient when scanning, as even slight body movements when tapping the screen and taking the eyes of the scanner made users more prone to mistakes. It was also utilized on other screens throughout the app, such as the exam protocol screen shown below.

    Scanning screen
    Exam protocol screen

    Wireflow

    A holistic visual representation of the user flow

    With ironed-out key screens, I mapped out a mid-fidelity wireflow to document the entire experience by combining wireframes and flowcharts to show complete views, paths, and interactions between screens.
    ‍
    Wireflows allowed us to have a common understanding of the product structure, supported me in gathering stakeholder feedback, and facilitated design discussions. As a result, Wireflows became an essential living document and were heavily referenced by the engineering team during implementation.

    User Testing

    Evaluating the design at the different stages

    I tested at different stages of the design process. For example, I sometimes tested early to validate or invalidate assumptions before spending engineering time and resources. Still, at other times I would test before a release roll-out to see if specific elements or flows needed improvements. The approach was a mixed bag. It ranged from guerrilla-style testing to more formal in-person testing with relevant participants in the medical space.
    ‍
    We also observed real users interact with the actual product and gather key qualitative insights during clinical trial training sessions with nurses.

    Testing session with a medical professional

    Design Library & handoff

    Creating a visual language

    The visual design was created from scratch, opting for a dark mode to align with the dark ultrasound images. While I didn't need a design library right away, I created a basic one that grew over time to maintain consistency across the app and desktop web app. This also helped in organizing UI components into a single source of truth file for a unified design, quicker handoff, and smoother workflow. The design library includes logos, a color palette, 70+ UI components, icons, surfaces, typography, and stroke styles.

    Final designs

    Final Android app product shots
    Welcome
    Benefit 3
    Log in
    Interstitial
    Home
    Patient details
    Exam protocol
    Scanning
    Record compression
    Compression review
    End of exam
    Interactive training
    Video

    Outcomes

    ✓

    Delivered design on time

    The project's biggest constraint was time due to commitments with hospitals to begin clinical trials. However, I successfully managed to design, test, and deliver the final product on time despite an ambitious deadline.

    ✓

    Software deployment in 7 Hospitals

    AutoDVT is already live in 7 hospitals in the UK and is in discussions to be implemented in further hospitals across Europe, Canada, and the United States.

    ✓

    Favorable preliminary results

    Preliminary results indicate AutoDVT to be 30x faster than the existing clinical pathway without sacrificing accuracy.

    ✓

    Established a relationship

    Through my work and collaboration, I gained the team's trust, which led to additional work, such as researching the feasibility of future applications, designing the company's marketing website, and designing the reviewer's Android mobile app.

    Next

    Android App

    In-app ultrasound training

    Achieved 83% task completion rate by 3rd practice exam by delivering 1st ever self-guided training for nurses—replaced in person training and established scalable training alternative

    I'm open to product design opportunities that deliver meaningful impact.

    Get in touch
    LINKEDIN
    COPY MY EMAIL
    COPIED TO CLIPBOARD