Code for "Steerable Pyramid Transform Enables Robust Left Ventricle Quantification"

This is an end-to-end framework for accurate and robust left ventricleindices quantification, including cavity and myocardiumareas, six regional wall thicknesses, and three directional dimensions.

The proposed method first decomposes a cardiovascular magnetic resonance image into directional frequency bands via Steerable Pyramid Transformation. Then deep representations of each direction are extracted separately via a CNN model and the temporal correlation between frames were modeled with a recurrent neural network. Finally, we explore the multidirectional relationship of features, indices, and directional subbands to optimize the quantification system.

The whole framework is shown below:

The dataset we used can be found at the MICCAI 2018/2019 Left Ventricle Full Quantification Challenge, an open source dataset on Kaggle.