A5029291582- Advanced Neural Network Applications
- Cell Image Analysis Techniques
- Advanced Image and Video Retrieval Techniques
- Medical Image Segmentation Techniques
- Remote Sensing and LiDAR Applications
- Single-cell and spatial transcriptomics
- Human Mobility and Location-Based Analysis
- Smart Agriculture and AI
- Advanced Clustering Algorithms Research
- Neural dynamics and brain function
- Spatial and Panel Data Analysis
- Metabolomics and Mass Spectrometry Studies
- Leaf Properties and Growth Measurement
- Impact of Light on Environment and Health
- Gene expression and cancer classification
- Galaxies: Formation, Evolution, Phenomena
- Astronomy and Astrophysical Research
- Mass Spectrometry Techniques and Applications
- Machine Learning and Data Classification
- Economics of Agriculture and Food Markets
- Advanced Fluorescence Microscopy Techniques
- Geophysics and Gravity Measurements
- Mobile Crowdsensing and Crowdsourcing
- Data Management and Algorithms
- Video Surveillance and Tracking Methods
European Molecular Biology Organization
2022
Heidelberg University
2017-2021
Heidelberg University
2019
Quantitative analysis of plant and animal morphogenesis requires accurate segmentation individual cells in volumetric images growing organs. In the last years, deep learning has provided robust automated algorithms that approach human performance, with applications to bio-image now starting emerge. Here, we present PlantSeg, a pipeline for tissues into cells. PlantSeg employs convolutional neural network predict cell boundaries graph partitioning segment based on predictions. was trained...
Image partitioning, or segmentation without semantics, is the task of decomposing an image into distinct segments, equivalently to detect closed contours. Most prior work either requires seeds, one per segment; a threshold; formulates as multicut / correlation clustering, NP-hard problem. Here, we propose efficient algorithm for graph "Mutex Watershed''. Unlike seeded watershed, can accommodate not only attractive but also repulsive cues, allowing it find previously unspecified number...
Abstract Spatial metabolomics using imaging mass spectrometry (MS) enables untargeted and label-free metabolite mapping in biological samples. Despite the range of available MS protocols technologies, our understanding detection under specific conditions is limited due to sparse empirical data predictive theories. Consequently, challenges persist designing new experiments, accurately annotating interpreting data. In this study, we systematically measured detectability 172...
The Fisher matrix is a widely used tool to forecast the performance of future experiments and approximate likelihood large data sets. Most forecasts for cosmological parameters in galaxy clustering studies rely on approach large-scale like DES, Euclid, or SKA. Here we improve upon standard method by taking into account three effects: finite window function, correlation between redshift bins, uncertainty bin redshift. first two effects are negligible only limit infinite surveys. third effect,...
We propose a theoretical framework that generalizes simple and fast algorithms for hierarchical agglomerative clustering to weighted graphs with both attractive repulsive interactions between the nodes. This defines GASP, Generalized Algorithm Signed graph Partitioning <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sup> Code available at: https://github.com/abailoni/GASP, allows us explore many combinations of different linkage criteria...
Summary Single-cell metabolomics promises to resolve metabolic cellular heterogeneity, yet current methods struggle with detecting small molecules, throughput, and reproducibility. Addressing these gaps, we developed HT SpaceM, a high-throughput single-cell method novel cell preparation, custom glass slides, small-molecule MALDI imaging mass spectrometry protocol, batch processing. We propose unified framework covering essential data analysis steps including quality control,...
ABSTRACT Quantitative analysis of plant and animal morphogenesis requires accurate segmentation individual cells in volumetric images growing organs. In the last years, deep learning has provided robust automated algorithms that approach human performance, with applications to bio-image now starting emerge. Here, we present PlantSeg, a pipeline for tissues into cells. PlantSeg employs convolutional neural network predict cell boundaries graph partitioning segment based on predictions. was...
We propose a theoretical framework that generalizes simple and fast algorithms for hierarchical agglomerative clustering to weighted graphs with both attractive repulsive interactions between the nodes. This defines GASP, Generalized Algorithm Signed graph Partitioning, allows us explore many combinations of different linkage criteria cannot-link constraints. prove equivalence existing methods some those introduce new have not been studied before. study empirical properties these define an...
Semantic instance segmentation is the task of simultaneously partitioning an image into distinct segments while associating each pixel with a class label. In commonly used pipelines, and label assignment are solved separately since joint optimization computationally expensive. We propose greedy algorithm for graph labeling derived from efficient Mutex Watershed algorithm. It optimizes objective function closely related to Symmetric Multiway Cut empirically shows scaling behavior. Due...
Calcium imaging is one of the most important tools in neurophysiology as it enables observation neuronal activity for hundreds cells parallel and at single-cell resolution. In order to use data gained with calcium imaging, necessary extract individual their from recordings. We present DISCo, a novel approach cell segmentation videos. temporal information recordings computationally efficient way by computing correlations between pixels combine shape-based identify active well non-active...
This work introduces a new proposal-free instance segmentation method that builds on single-instance masks predicted across the entire image in sliding window style. In contrast to related approaches, our concurrently predicts all masks, one for each pixel, and thus resolves any conflict jointly image. Specifically, predictions from overlapping are combined into edge weights of signed graph is subsequently partitioned obtain final instances concurrently. The result parameter-free strongly...