publications

2024

1. 

   ICML Topological Deep Learning Challenge 2024: Beyond the Graph Domain

   Guillermo Bernárdez , Lev Telyatnikov , Marco Montagna , and 70 more authors

   2024

   Abs PDF Code

   This paper describes the 2nd edition of the ICML Topological Deep Learning Challenge that was hosted within the ICML 2024 ELLIS Workshop on Geometry-grounded Representation Learning and Generative Modeling (GRaM). The challenge focused on the problem of representing data in different discrete topological domains in order to bridge the gap between Topological Deep Learning (TDL) and other types of structured datasets (e.g. point clouds, graphs). Specifically, participants were asked to design and implement topological liftings, i.e. mappings between different data structures and topological domains –like hypergraphs, or simplicial/cell/combinatorial complexes. The challenge received 52 submissions satisfying all the requirements. This paper introduces the main scope of the challenge, and summarizes the main results and findings.

2023

1. 

   CLEM-Reg: An automated point cloud based registration algorithm for correlative light and volume electron microscopy

   Daniel Krentzel , Matouš Elphick , Marie-Charlotte Domart , and 5 more authors

   May 2023

   Abs PDF Code

   Correlative light and volume electron microscopy (vCLEM) is a powerful imaging technique that enables visualisation of fluorescently labelled proteins within their ultrastructural context on a subcellular level. Currently, expert microscopists find the alignment between acquisitions by manually placing landmarks on structures that can be recognised in both imaging modalities. The manual nature of the process severely impacts throughput and may introduce bias. This paper presents CLEM-Reg, a workflow that automates the alignment of vCLEM datasets by leveraging point cloud based registration techniques. Point clouds are obtained by segmenting internal landmarks, such as mitochondria, through a pattern recognition approach that includes machine-learning. CLEM-Reg is a fully automated and reproducible vCLEM alignment workflow that requires no prior expert knowledge. When benchmarked against experts on two newly acquired vCLEM datasets, CLEM-Reg achieves near expert-level registration performance. The datasets are made available in the EMPIAR public image archive for reuse in testing and developing multimodal registration algorithms by the wider community. A napari plugin integrating the algorithm is also provided to aid adoption by end-users. The source-code for CLEM-Reg and installation instructions can be found at https://github.com/krentzd/napari-clemreg .