Echonet

• 

  Poster

Inspiration

Paper from Stanford University (origin of the database).

What it does

Estimate Ejection Fraction from cardiac ultrasound videos by segmenting the left ventricle in the frames corresponding to end systole and end diastole.

How we built it

One kind of CNN architecture called U-Net, very used in healthcare applications of deep learning.

Challenges we ran into

The given database was really difficult to work with, especially the labels. Instead of segmentation masks for the images, we were given corners of the rectangles in which the segmentation could be outlined.

Accomplishments that we're proud of

What we learned

What's next for Echonet

Built With

• python
• tensorflow