IBM: Watson for Oncology



Watson For Oncology Marketing Website


What is Watson for Oncology?

Watson for oncology is a web-based application that will allow oncologists to search an immense set of oncological data quickly from one of the top trusted research institutions in the United States, Memorial Sloan Kettering. With the help of NLP and some AI magic, an oncologist can search a wide variety of medical treatments for their patients in seconds.

Watson for Oncology is being tested and improved in multiple hospitals partnered with Memorial Sloan Kettering in the United States and Canada.




The Problem

There is an enormous amount of research and data collected to help find treatments for various types of cancer. Most of the data today is stored in databases, journals, books, etc. Hospitals operate with antiquated systems to store patient data. These systems are called EMRs or Electronic Medical Records. Some patients have paper copies of their treatments, medication information, and other patient information that has to be translated into the EMR. This “lose leaf” data the patient brings, is unstructured.



What We Set Out to Do

We set out to build a product that could bring all of the best research and unstructured patient data together in order to help oncologists search, analyze, and decipher massive amounts of data.

When I joined the team, a beta version of the application existed. It was part of my responsibility to validate the beta with our users and make recommendations to improve the experience based off of their feedback.


"We set out to build a product that could bring all of the best research and unstructured patient data together"







Research & Understanding

Before I joined the team, generative research was conducted around the world of oncology and how oncologists make their decisions. Part of my role was to reevaluate our findings and take a deeper look at our users. This involved in-person interviews, journey maps, constructing personas, and analysis.


The People

Our target users are typically teams of about 2 - 4 oncologists per team. This can vary, however, based on the type of institution you target. Smaller oncology centers that focus on treatment in a community (community oncologist/general oncologist) versus a larger research institution that can focus more on training, study and treatment. While team size can vary, the common thread is that each oncologist generally has a specialty. *Specialty: type of cancer they focus their attention on. 8 + years of education > residency > training with specialist.

Their Motivations

  • #1 focus is their patients care and comfort during their treatment.

  • The pursuit for finding ways to cure or manage their patient's disease.

  • Advancing the field of oncology through research and study.

  • Knowledge sharing in addition to advancing research and documenting and publishing their knowledge is important for the oncology community.


Beta Testing

When I joined the team, a beta version of the application existed. It was part of my responsibility to validate the beta with our users and make recommendations to improve the experience based off of the feedback.



What We Discovered

This tool is excellent for research and training. It provided ways for the oncologist to input different or incorrect values to test the outcomes. In some ways, they used Watson as a forecasting tool.

The findings showed that the product as whole was able to determine fairly accurate treatment plans for patients. However, even though the tool functioned properly, it did not provide an incredible amount of value. It wasn’t telling the oncologist information they didn’t already know.



Quotes from Oncologists

“We train our fellows in residence to question everything when they are presented with data. In an evidence-based practice, Watson delivery a result with a green check next to it isn’t good enough.”

“If I’m teaching residents that are going by case-based teaching, then I could use Watson to say "OK-let's review this case and we can go through the different scenarios". I could modify the receptors to look at the risk and benefits to see what the recommendations are.”







“even though the tool functioned properly, it did not provide an incredible amount of value.”





Insights we Learned at this Stage

Oncology is an evidence-based practice. Hypotheses are created, then tested. An example of that is clinical trials. These findings become part of journals, case studies, and published papers. Sometimes these findings eventually become part of everyday medicine.

Recommendations Moving Forward

As a team, our findings began to build an experience around training and particle use in day to day practice.
  • Fluid EMR integration.

  • Patient clustering by illness or case type for studies.

  • Creating dynamic layouts for exploring content.

  • Creating more efficient ways to manipulate attributes

  • Ability to export data and treatment options for research studies.



"Oncology is an evidence-based practice. Hypotheses are created, then tested."




Design and Concept Development

During the exploration phase, the team and I went through various story boards, scenarios, and use cases associated with our insights. We compared the data we collected to our overall goal and assumptions we made early on.

We began to rebuild our experience around the idea that we could pivot to create a powerful research tool that would give our user the flexibility to change attributes quickly and process data rapidly.






High-Level Architecture

Below is a condensed version of the IA proposed for updating the existing experience. It includes 3 core screens that the UX updates touch the most.










Recommended Enhancements

The new UX follows a 3 step process for digesting and understanding patient data, as well as navigating large amounts of data retuned on treatments plans.





Lets talk...

Want to learn more about this project? Send me an
email and lets talk!