Our project has the objective of improving the training efficiency and overall accuracy of video-based cardiovascular disease detection models, specifically using echocardiograms. We initially hoped to create a self-supervised model that will outperform the fully supervised model by using x3d architecture and echonet dynamic dataset as our video data. The project itself has the potential of leaving an impact on the overall machine learning based disease detection study as it will allow various disease detection models to effectively learn even though there is a scarcity of labeled videos. These disease detection models are expected to assist healthcare professionals, and minimize human error when it comes to detecting and diagnosing various diseases.
