Echo-SyncNet:

A Neural Network for Synchronization of Cardiac Echo

Example synchronization of three unique cardiac view angles.

The need for automatic echo synchronization:

• Calculation of clinical measurements in cardiac echo often require or benefit from having multiple synchronized views or accurately annotated keyframes.
• Traditional methods to synchronize echo rely on external factors such as an electrocardiogram which may not always be available; especially in the point of care setting.

To address these points we propose Echo-SynNet a neural network-based framework for automatic synchronization of cardiac echo. Echo-SynNet is trained using only self-supervised methods and is hence cheap to train or finetune on any dataset.

Echo-SyncNet is an encoder style CNN trained to produced low dimensional and feature-rich embedding sequences cardiac ultrasound videos. The embedding vectors carry a powerful semantic understanding of the structure and phase of the heartbeat. Videos can be synchronized simply by performing feature matching on their embedding sequences.

---

Echo-SyncNet is trained on a dataset of 3070 unannotated echo studies. We use a multiobjective self-supervised loss, described in detail in our paper, to promote the consistency of embedding features across multiple training samples.

Notebook Demo: Coming Soon!