Project 1.1 - High Dimensional Data
We develop deep learning techniques and efficient architectures for quantitative analysis of 4- and 5-D medical images that make optimal use of additional dimensions and apply them to cardiac spectral CT and MRI and sequential 4D chest CT.
Project 1.2 - Deep Generative Models
In medical applications large data sets are generally not available, or they are unbalanced (containing much fewer abnormal, e.g. cancer cases, than normal data). We address the challenge of simulating abnormal training data by generative deep learning. The new technology is based on variational autoencoders, with approximate Bayesian computation and extensions for semi-supervised learning.
Project 1.3 - Deep Transfer Learning
For real world systems training data has been acquired with slightly different acquisition protocols, different scanners, or from a different patient population. To still learn robustly, we will develop deep transfer learning technology where the domain transfer is addressed in the representation learning step for which we use different coupled network architectures.
Project 2.1 - Weakly Labeled Learning
We focus on deep learning with large amounts of images that are only weakly labeled (e.g. only overall diagnosis or treatment outcome is available). We develop techniques that exploit a small set of images with detailed annotations and a large pool of weakly or completely unlabeled data. We exploit shared representations between learning tasks with different localization levels and use active learning where medical experts are asked for feedback on automatically selected cases.
Project 2.2 - Dynamic Deep Learning
We develop a framework for dynamic deep learning. This system can learn continuously from feedback from experts. It can learn easy concepts first and gradually learn complex tasks after having seen more data. The network will express its uncertainty and ask for feedback on cases it is uncertain about or has not seen before.