Building AI trust: iKure + The IBM Data Science and AI Elite team tackle bias to improve healthcare outcomes
With only one cardiologist for every 6,000 patients, there are not enough specialists to address India’s most pressing health concern: heart disease.
Leaders at Kolkata-based health tech company iKure realized they could fill that critical gap using wearable technology and machine learning. The start-up worked with the IBM Data Science and AI Elite team (DSE) to create a model within IBM Cloud Pak for Data that could identify patients who have the highest risk of suffering from a heart attack, allowing doctors to see the most urgent cases first, and ultimately save lives.
Check out this video to learn more about the DSE demonstrating how Cloud Pak for Data can bring the power of AI to iKure’s data stored in an AWS, MySQL data store.
When iKure’s CEO and founder Sujay Santra logs onto to Cloud Pak for Data, he can easily check on the performance of his new model. In a matter of minutes, he can harness the capabilities of Watson OpenScale to detect bias and predict whether he can trust his AI model, all without relying on a team of data scientists to do it for him.
For example, his dashboard displays a red “bias” alert – since male patients were flagged as having a higher rate of cardiovascular disease than female patients. The patient data used to train the model might have a problem since it had triple the number of male patient data. But could there be other reasons for the bias? Sujay decided to compare his data to a published risk model. The search field pulled up a healthcare accelerator, which acts as a ladder to AI jumpstart kit for workflow. He was able to easily evaluate the bias detected from his training data and select whether to keep it or remove it.
iKure can access and combine patient population data from its mobile healthcare information management system with EKG signal data, can classify the EKG data as normal or with anomalies -- and then present a prioritized list of patient cases to the call center physician for review based on potential acuteness via a cloud-based web application.
Standardizing on Cloud Pak for Data allows companies like iKure to get up and running quickly without needing to recode if they have to deploy or embed their application elsewhere. This is especially valuable for companies that need to be agile no matter what cloud technology they have today or tomorrow.
“IBM Cloud Private for Data facilitated the model development and deployment of a predictive model for cardiac care for iKure,” said Sujay Santra, iKure’s CEO and founder. “IBM’s Data Science Elite team demonstrated the model development process in ICP for Data via Watson Studio with multiple AWS data sources. It also proved model accuracy using patient clinical and demographic variables and physician feedback with the added benefits of rapid model development, publication and iteration.”