We focus on research projects in the fields of:
Identifying similar objects in a data set is critical for the efficient use of machine learning techniques. Because the number of pairwise similarities grows quadratically in the data set size, this has proven to be a complex problem. We introduce a sparse computation method that benefits from focusing only on relevant similarities. For data sets with up to 8.5 million objects, we show that the novel approach significantly improves running times with minimal loss in accuracy.
Read our TBD-paper to learn more.
Kidney-related diseases are a major global health issue. An important indicator for determining the health state of a kidney is the number and volume of glomeruli (units responsible for blood filtration). This article proposes a semi-supervised learning approach for the large-scale glomeruli analysis of micro-computed tomography (CT) images. We demonstrate the ability of the approach to provide accurate estimates of glomeruli counts and volume at different levels of image coverage.
Read our MICCAI-paper to learn more.
