Echocardiograms are ultrasound images of the heart, that are lower in cost and quicker to perform than other imaging techniques such as MR images and CT scans. They are thus the most frequently used cardiac imaging technique for preventative screenings and ongoing monitoring of heart conditions. Cardiologists spend a significant amount of time analysing echocardiograms and reporting on the results. Many of these analytical studies are for people with normal heart functioning that require no further medical intervention. Automating the identification of normal heart function from echocardiogram images can potentially reduce the time that cardiologists spend in front of computers and help them increase the amount of time spent with ill patients that need them most.
The Deep Learning for Echocardiogram Analysis code base was developed and open-sourced during the 2019 Data Science for Social Good Fellowship programme. The project automates the analysis of echocardiogram images to detect normal heart functioning.
2019: Responsible for end-to-end data pipeline design