Jupyter Notebook, describing the data preprocessing used for the fourth project of the "Advanced Machine Learning" course from ETH.
Non-negative matrix factorization was used to obtain 'eigen-images' of the videos. All the frames of the new generated videos are a linear combination of that "eigen-images" which in fact correspond to the end-systole and end-diastole of a cardiac cycle.
Machine Learning for Cardiac Ultrasound Time Series Data