Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Stroke Recognition using β-Variational Autoencoder
develop and implement a β-Variational Autoencoder to disentangle stroke lesions features
Keywords: Keywords: Magnetic Resonance imaging, MRI, stroke, machine learning, deep neural network, convolutional neural network, variational autoencoders, VAE.
An ischemic stroke is a sudden loss of neurological function due to brain parenchyma damage caused by the sudden loss or significant decrease of blood flow to a specific region of the brain. Stroke is the second most common cause of death worldwide, the leading cause of acquired disability in adults, and the third cause of loss of years of life.
Deep neural networks have been very successful at automatic extraction of meaningful features from data. In particular Variational Auto-Encoders (VAE) are generative model that can disentangle simple data features from a highly complex input space by factorizing a latent representation1. β-VAE2 are an extension of VAE by introducing an adjustable hyperparameter beta that balances latent channel capacity and independence constraints to improve reconstruction accuracy.
1. D. Kingma, M. Welling. Auto-Encoding Variational Bayes, Proceedings of the 2nd International Conference on Learning Representations (ICLR), 2014
2. I. Higgins et al. beta-VAE: Learning Basic Visual Concepts with a Constrained Variational Framework, https://openreview.net/forum?id=Sy2fzU9gl
An ischemic stroke is a sudden loss of neurological function due to brain parenchyma damage caused by the sudden loss or significant decrease of blood flow to a specific region of the brain. Stroke is the second most common cause of death worldwide, the leading cause of acquired disability in adults, and the third cause of loss of years of life.
Deep neural networks have been very successful at automatic extraction of meaningful features from data. In particular Variational Auto-Encoders (VAE) are generative model that can disentangle simple data features from a highly complex input space by factorizing a latent representation1. β-VAE2 are an extension of VAE by introducing an adjustable hyperparameter beta that balances latent channel capacity and independence constraints to improve reconstruction accuracy.
1. D. Kingma, M. Welling. Auto-Encoding Variational Bayes, Proceedings of the 2nd International Conference on Learning Representations (ICLR), 2014 2. I. Higgins et al. beta-VAE: Learning Basic Visual Concepts with a Constrained Variational Framework, https://openreview.net/forum?id=Sy2fzU9gl
The objective of this project is to develop and implement a β-Variational Autoencoder to disentangle stroke lesions features using a large (>500 cases) of strokes diffusion-weighted magnetic resonance images.
Programming experience in Python, C/C++ or equivalent is required.
The objective of this project is to develop and implement a β-Variational Autoencoder to disentangle stroke lesions features using a large (>500 cases) of strokes diffusion-weighted magnetic resonance images.
Programming experience in Python, C/C++ or equivalent is required.
Dr. Christian Federau, federau@biomed.ee.ethz.ch)
Supervising Professor: Prof. Dr. Sebastian Kozerke (kozerke@biomed.ee.ethz.ch)
Dr. Christian Federau, federau@biomed.ee.ethz.ch) Supervising Professor: Prof. Dr. Sebastian Kozerke (kozerke@biomed.ee.ethz.ch)