Kardia Visus

This case study demonstrates my learnings on how to successfully navigate the opportunities and challenges of a BigCo (Big Company) and build products at scale in complex environments along with 5 other students from Cornell Tech with diverse academic background. Under guidance of Johnson & Johnson Design Studio, our team built KARDIA VISUS.

Spring 2022 — 20 weeks

Skills: UX Research, Prototype, UI Design

Teammates: Adam Chew (CS), Gazi Inkiyad (ECE), Janet Liao (MBA), Sofia Thompson (ORIE), So Young Park (Design), Xingyu Tao (CS)

 

Company Analysis

Johnson & Johnson is a multinational corporation that enables better health experiences for diverse communities – consumers, employees, healthcare providers and patients. With such practice, the company primarily focuses on providing reliable and inclusive services in three separate areas – medical devices, pharmaceuticals, and consumer health. To visualize the company’s overview and its ecosystem, our team has created a rich picture.

 

USER RESEARCH

As we began to delve deeper to identify areas of opportunity, as a team, we conducted research and interviews primarily to identify and learn more about healthcare services in any possible segments. We set research goals with the following:

1.    To identify and learn more about healthcare services in any possible segments

2.    To find public health intervention and the consequential risk factors

 

USER INTERVIEWs

We conducted interviews in three different segments:

  1. Consumers – to understand direct to consumer mode

  2. Medical professionals – to understand any emerging technical development in medical field that we can access public dataset to test extensively

  3. Patients who regularly meet with their assigned doctors – to expand our scope to help addressing some of the toughest health challenges people face (How do we transform healthcare system from diagnose and treat to predict and prevent?)

    • We found these group of interviewees on online community who were active and willing to talk about themselves because we wanted to be aware that sometimes talking about their illness can be difficult

 

AREAS OF OPPORTUNITY

Here were 7 areas of opportunity highlighted in our interviews:

On top of this, we realized that there are many platforms and application that benefits patients such as ZocDoc. However, there are not enough advancement that will enhance workflow of doctors or other medical partitioners. Therefore, we wanted to create a solution that will enhance both doctors and patient’s experience because they should benefit simultaneously.

 

THE 3 IDEAS

With our company advisor, we narrowed down our ideas to three. We created product narrative and business canvas model to discuss more in-depth for its potential and liability of each idea.

 

Note taking devices 

  • Saving time for healthcare providers with a personalized note taking device that accurately transcribes medical terms and predicts potential diagnosis results

  • This seemed to be the furthest from J&J business model as their revenue is based on making and selling products and services. Our company advisor believed that it will be only applicable in surgical setting.

  • Quickly and better understand patients

Preventative Care

  • Increase preventative care usage by helping patient to understand and recognize the importance of medication adherence

  • Medical adherence is a huge problem for J&J and globally. We wanted to work towards application to provide clear instructions about what patients should be doing or providing prescription info that addresses chronic conditions such as myopia

Cardiac Mapping Tool

  • Assisting cardiologists and surgeons by visualizing affected heart regions from ECG signals to increase confidence in the diagnosis

 

Detailed Value Creation Analysis

Product Narrative

 

Business Model Canvas

We created Business Model Canvas to imagine a new venture — to add more detail to our ideas. We explored various hypotheses and variants of our idea to further iterate on with more research.

 

Currently in industry

Doctors already go through a consensus-driven diagnostic process for reading ECG signals, linking ECG patterns to different cardiac issues is a well researched area of cardiology. This could be time-consuming when the patient is in critical condition.

What Already Exists:

  • Doctors already go through a consensus-driven diagnostic process for reading ECG signals

  • Linking ECG patterns to different cardiac issues is a well researched area of cardiology

Our Solution

Doctors could save time before, during and even after surgery by being able to directly see a visualization of the potential heart area being affected. Saving time, and saving lives.

Differentiating Aspects of This Product:

  • AI model assisted labeling and visualization of ECG could lead to more accurate and speedy diagnostics

  • Visualizing the area of the heart being potentially affected is not included in the majority of existing research, so this could be a differentiating and advantageous feature

Demand: Will people buy it?

We investigated demand side and identified potential customers segments who will find this visualization tool useful.

  1. The procurement decision makers for hospital systems and cardiac surgeons

    • Research done by the University of California San Francisco shows that physicians could benefit from enhanced ECG diagnosis ability offered by AI analysis

    • Electrocardiograms (ECGs) are the most common cardiovascular test worldwide. Millions of clinicians rely every day on automated preliminary ECG interpretation to assist with a wide range of cardiac diseases from urgent heart attacks to abnormalities of cardiac rhythm, electrical conduction or structure.

  2. Students in medical school for training purpose. There are different tiers in revenue stream and there is potential to lower the cost by offloading on one end

 

Supply Side: Can we make it?

  • There are existing trove of research papers on the topic of AI enhanced cardiography, such as this proof of concept published in the Nature journal.

  • There are multiple projects, such as this one done by General Electric, currently still in the research phase. So, this idea presents a large degree of potential benefit for the first system that could be commercially deployed in the market (hospital systems)

What we need:

  1. Dataset for Prototyping

    • ECG data with heart region labels is widely available on the internet for public use.

    • This includes ECG data to general cardiac abnormality "ground truth" dataset. For our testing, we primarily focused on Arrhythmia – improper beating of the heart

  2. 3D Modeling

    • Using existing ML and 3D modeling technologies, we can effectively model the decision process that surgeons have to go through in real time with the added benefit of a predictive model to improve the accuracy of their decisions.

    • On the front end design of the product, we may need to partner with someone who specializes in 3D modeling. A large part of our value proposition is that we hope to display real time images (maps) with prediction values for use during surgery.

PRoduct Validation

To validate if this product will be viable and useful to medical professionals – especially for cardiologists and electro physicians, we designed preliminary prototype to test user engagement and demand-side. We also conducted user interviews to test the interest and to see if there is any potential business opportunity in surgical field for J&J.

 

LOW-fi Prototype

Using low-fi prototype, we asked medical professionals to identify the problem and the affected region. The condition we focused for this exercise was premature ventricular contraction.

Survey A: Displaying two regular ECG signal to determine how long it took to interpret information from the graphs

Survey B: Displaying the problematic region to determine how much faster they were able to absorb the information with visualization tool

 

Participant Distribution

Our participant involved with doctors, student, and variety of medical professionals – including healthcare nurse, clinical specialist, physician assistant and even engineers and scientist from J&J medical device team.

 

TIME 

Based on the results that we collected in this survey, it showed that with the visualization support, the participants were able to answer the questions 30 to 60 seconds faster compared to having just ECG graphs.

 

Question Accuracy for Q1

The first question was testing to identify RVOT condition. In terms of answer accuracy, most of the participants could identify the problem and the location. The accuracy improved nearly 5% and reduced 60 seconds to answer correctly when they were presented with the visualization tool.

 

Question Accuracy for Q2

The answer to the second question was normal condition. In this case, we saw participants answered better without visualization. The result brought to our attention that our visual UI may not have been clear for participants to understand. Consequently, for the future experiment, we decided to further explore different UI design directions for clear communication.

 

USABiLITY Testing

After getting validation of our product, we continued to make iteration from low-fi prototype and found areas of improvement. We showed users two different UI designs to test which layout of information is the clearest and which features are essential or needed. This allowed us to understand which area lacks affordance and signifiers in user flow and user experience.

 

SURVEY DESIGN 

Hi-Fi Prototype

As we are trying to move forward with heat-map, we wanted to ensure that all the users will be aligned with what each color signifies because sometimes, colors can be very subjective and may easily lead to miscommunication. Therefore, we dedicated an area for users to refer to each color and what each color represents.

Link to our live demo

QR code to our live demo

NExt Step:

Even though our semester has ended, we are continuing this project, collaborating with Johnson and Johnson with a potential of sponsoring academic research, under guidance of Deborah Estrin, a professor in Computer Science at Cornell Tech program. We will continue to test concept feasibility which includes assessing global regulations as we continue to iterate while understanding more technical knowledge. Also, we wanted to embed more features that will enhance user’s experience.