1

Training state-of-the-art pathology foundation models with orders of magnitude less data
We present novel pathology FMs trained with an adaptation of DINOv2 and post-training on higher-resolution images, achieving superior or comparable performance to state-of-the-art FMs while using up to two orders of magnitude fewer WSIs.
Mikhail Karasikov, Joost van Doorn, Nicolas Känzig, Melis Erdal Cesur, Hugo Mark Horlings, Robert Berke, Fei Tang, Sebastian Otálora
Cite Project DOI Preprint Code Model
Training state-of-the-art pathology foundation models with orders of magnitude less data
Lossless Indexing with Counting de Bruijn Graphs
We propose the concept of Counting de Bruijn graphs generalizing the notion of annotated (or colored) de Bruijn graphs. Counting de Bruijn graphs supplement each node-label relation with one or many attributes (e.g., a k-mer count or its positions in genome).
Mikhail Karasikov, Harun Mustafa, Gunnar Rätsch, André Kahles
Cite Project DOI Preprint Code
Lossless Indexing with Counting de Bruijn Graphs
Topology-based Sparsification of Graph Annotations
We present RowDiff, a new technique for compacting graph annotations by leveraging expected similarities in labelings of adjacent nodes.
Daniel Danciu, Mikhail Karasikov, Harun Mustafa, André Kahles, Gunnar Rätsch
PDF Cite Project Slides DOI Code
Topology-based Sparsification of Graph Annotations
Sparse Binary Relation Representations for Genome Graph Annotation
We present Multi-BRWT, a scheme for compressed representation of sparse binary bitrices adaptive to different kinds of input data.
Mikhail Karasikov, Harun Mustafa, Amir Joudaki, Sara Javadzadeh No, Gunnar Rätsch, André Kahles
PDF Cite Project Slides DOI Code
Sparse Binary Relation Representations for Genome Graph Annotation