My work focuses on machine learning for medical image analysis. Machine learning algorithms learn from examples in order to make predictions about novel data. For example, by learning from medical scans with annotated abnormalities, the algorithms can detect abnormalities in previously unseen patients. One example is detecting the lung disease COPD. Because obtaining ground-truth annotations is time-consuming, I am in particular interested in learning scenarios where few labels are available, such as:

  • multiple instance learning
  • transfer learning or domain adaptation
  • crowdsourcing
Early experiences with crowdsourcing airway annotations in chest CT
The similarity between dissimilarities
Asymmetric similarity-weighted ensembles for image segmentation
Label stability in multiple instance learning
Dissimilarity-based ensembles for multiple instance learning
Single- vs. multiple-instance classification
On classification with bags, groups and sets
Characterizing multiple instance datasets
Multiple instance learning with bag dissimilarities
Dissimilarity-based multiple instance learning
Network-guided group feature selection for classification of autism spectrum disorder
Classification of COPD with multiple instance learning
On the informativeness of asymmetric dissimilarities
Combining instance information to classify bags
Bridging structure and feature representations in graph matching
Class-dependent dissimilarity measures for multiple instance learning
Does one rotten apple spoil the whole barrel?
Random subspace method for one-class classifiers
Bag dissimilarities for multiple instance learning
Pruned random subspace method for one-class classifiers