IBM jStart enables MD Anderson Cancer Center with Enhanced Cohort Selection Capabilities by Harnessing Unstructured Clinical Information.

mdanderson building
What started as an IBM jStart Proof of Concept (PoC) for MD Anderson’s Chief Medical Information Officer, Dr. Frenzel, blossomed into a complete revelation in the way reseachers and clinicians utilize historical patient experiences. Mining the data from hundreds and even thousands of patient records to cull treatment patterns and success rates would be a nearly impossible human feat. Leveraging IBM Watson Content Analytics (WCA) to automate the extraction of breast cancer information from clinical documentation seemed daunting, at best. Previously, clinicians logged 1020 hours to complete the chart conversion for a mere 3000 patients. jStart was able to exploit the Watson Technologies within 240 hours thus creating an app that would converted the same 3000 patient files in minutes. The information extracted, leveraging Natural Language Processing (NLP), contained straightforward medical concepts but also abstracted concepts like Satisfaction Indicators of Cosmetic Surgery results. The POC was a smashing success.

DrFrenzelThis led to a follow-on project for the IBM Watson team, who then designed and integrated WCA capabilities into the MD Anderson Oncology Expert Advisor (OEA) solution powered by IBM Watson. Researchers were in need of synthesized Leukemia patient clinical information to create actionable decision-trees for newly diagnosed patients.

With the success of the MD Anderson OEA work, Dr. Frenzel wanted to bring the proven NLP capabilities of WCA for use across all of MD Anderson departments and for all clinicians. He wanted to establish the infrastructure and enable MD Anderson skills to allow for the population of unstructured clinical information into their clinical data warehouse.

The first application to exploit newly added clinical information was targeted to improve their cohort selection capabilities. Dr. Frenzel once again engaged jStart to work with MD Anderson technical staff in order to architect and implement an integrated WCA infrastructure with their data warehouse and the cohort selection application. He also requested that we deploy the jStart NLP modeling methodology for skill mastery so that MD Anderson would be in a position to development specialized NLP extraction model going forward.

Today, MD Anderson has an NLP infrastructure that is feeding NLP derived data into the data warehouse and that data is leveraged by their cohort selection application. MD Anderson is at the end of the user acceptance validation for their first use-case where NLP models were solely implemented by MD Anderson NLP expertise.

In a short two years, MD Anderson has gained strong technical skills and expertise as a result of partnering with IBM jStart. Due to the unwavering commitment and support of Dr. Frenzel, MD Anderson has moved ever-closer to their vision:
“Allow our Clinicians, Scientist, Researchers to stand on the shoulders of everyone else in the practice and begin to see across patients, beyond the patients seen in their own experience… to allow them to look beyond the current patient and look at large cohorts of patients and begin to tease out those things that people have not put together before” – Dr John Frenzel, M.D.

IBM jStart develops unique solutions using Emerging Technologies.   See how jStart can help you.