Peter Hirsch
Research Subjects
Peter Hirsch has a total of 40 co-authors. Here are the top co-authors by number of shared publications:
- Dagmar Kainmueller
Max-Delbrueck-Center for Molecular Medicine in the Helmholtz Association (MDC)
6 shared publications - Lisa Mais
Max-Delbrueck-Center for Molecular Medicine in the Helmholtz Association (MDC)
4 shared publications - Jan Funke
Janelia Research Campus, Howard Hughes Medical Institute, Ashburn, VA, USA
3 shared publications - Stephan Preibisch
Janelia Research Campus, Howard Hughes Medical Institute, Ashburn, VA, USA
3 shared publications - Caroline Malin-Mayor
Janelia Research Campus, Howard Hughes Medical Institute, Ashburn, VA, USA
2 shared publications - Claire Managan
Janelia Research Campus, Howard Hughes Medical Institute
2 shared publications
Latest Publications
- Open Access
Neuron types and distribution of a Drosophila driver line collection
Cell type designations label uniquely-identifiable groups of related cells that typically function in similar ways. Cell types divide the Drosophila melanogaster central nervous system into functional units, but remain challenging to target and identify. Previously we generated a driver line and image resource for targeting many neuron types ( Meissner et al., 2025 ). Here we improve on identifying the targeted neurons by automated searching against electron microscopy datasets and manual scoring of search results to add 6,368 annotations across 1,772 driver lines in the collection. These annotations provide a potential training resource for future search algorithms. We then estimate the number of cell types and thus the fraction of the central nervous system labeled by the collection as a whole.
Open AccessNeuron types and distribution of a Drosophila driver line collection
Cell type designations label uniquely-identifiable groups of related cells that typically function in similar ways. Cell types divide the Drosophila melanogaster central nervous system into functional units, but remain challenging to target and identify. Previously we generated a driver line and image resource for targeting many neuron types ( Meissner et al., 2025 ). Here we improve on identifying the targeted neurons by automated searching against electron microscopy datasets and manual scoring of search results to add 6,368 annotations across 1,772 driver lines in the collection. These annotations provide a potential training resource for future search algorithms. We then estimate the number of cell types and thus the fraction of the central nervous system labeled by the collection as a whole.
- 30 September 2026
- Published by Springer Nature
- p. 502-519
Unsupervised Semantic Segmentation Facilitates Model Understanding
- 30 September 2026
- Published by Springer Nature
- p. 502-519
Unsupervised Semantic Segmentation Facilitates Model Understanding
- 16 September 2022
- Published by Springer Nature
- p. 25-35
Tracking by Weakly-Supervised Learning and Graph Optimization for Whole-Embryo C. elegans lineages
- 16 September 2022
- Published by Springer Nature
- p. 25-35
Tracking by Weakly-Supervised Learning and Graph Optimization for Whole-Embryo C. elegans lineages
- Open Access
- 5 September 2022
- Published by Springer Nature in Nature Biotechnology
- Vol. 41 (1) , 44-49
Automated reconstruction of whole-embryo cell lineages by learning from sparse annotations
We present a method to automatically identify and track nuclei in time-lapse microscopy recordings of entire developing embryos. The method combines deep learning and global optimization. On a mouse dataset, it reconstructs 75.8% of cell lineages spanning 1 h, as compared to 31.8% for the competing method. Our approach improves understanding of where and when cell fate decisions are made in developing embryos, tissues, and organs.
- 5 September 2022
- Published by Springer Nature in Nature Biotechnology
- Vol. 41 (1) , 44-49
Open AccessAutomated reconstruction of whole-embryo cell lineages by learning from sparse annotations
We present a method to automatically identify and track nuclei in time-lapse microscopy recordings of entire developing embryos. The method combines deep learning and global optimization. On a mouse dataset, it reconstructs 75.8% of cell lineages spanning 1 h, as compared to 31.8% for the competing method. Our approach improves understanding of where and when cell fate decisions are made in developing embryos, tissues, and organs.
- Open Access
Automated Reconstruction of Whole-Embryo Cell Lineages by Learning from Sparse Annotations
We present a method for automated nucleus identification and tracking in time-lapse microscopy recordings of entire developing embryos. Our method combines deep learning and global optimization to enable complete lineage reconstruction from sparse point annotations, and uses parallelization to process multi-terabyte light-sheet recordings, which we demonstrate on three common model organisms: mouse, zebrafish, Drosophila . On the most difficult dataset (mouse), our method correctly reconstructs 75.8% of cell lineages spanning 1 hour, compared to 31.8% for the previous state of the art, thus enabling biologists to determine where and when cell fate decisions are made in developing embryos, tissues, and organs.
Open AccessAutomated Reconstruction of Whole-Embryo Cell Lineages by Learning from Sparse Annotations
We present a method for automated nucleus identification and tracking in time-lapse microscopy recordings of entire developing embryos. Our method combines deep learning and global optimization to enable complete lineage reconstruction from sparse point annotations, and uses parallelization to process multi-terabyte light-sheet recordings, which we demonstrate on three common model organisms: mouse, zebrafish, Drosophila . On the most difficult dataset (mouse), our method correctly reconstructs 75.8% of cell lineages spanning 1 hour, compared to 31.8% for the previous state of the art, thus enabling biologists to determine where and when cell fate decisions are made in developing embryos, tissues, and organs.
- Open Access
PatchPerPixMatch for Automated 3d Search of Neuronal Morphologies in Light Microscopy
Studies of individual neurons in the Drosophila nervous system are facilitated by transgenic lines that sparsely and repeatably label respective neurons of interest. Sparsity can be enhanced by means of intersectional approaches like the split-GAL4 system, which labels the positive intersection of the expression patterns of two (denser) GAL4 lines. To this end, two GAL4 lines have to be identified as labelling a neuron of interest. Current approaches to tackling this task include visual inspection, as well as automated search in 2d projection images, of single cell multi-color flip-out (MCFO) acquisitions of GAL4 expression patterns. There is to date no automated method available that performs full 3d search in MCFO imagery of GAL4 lines, nor one that leverages automated reconstructions of the labelled neuron morphologies. To close this gap, we propose PatchPerPixMatch, a fully automated approach for finding a given neuron morphology in MCFO acquisitions of Gen1 GAL4 lines. PatchPerPixMatch performs automated instance segmentation of MCFO acquisitions, and subsequently searches for a target neuron morphology by minimizing an objective that aims at covering the target with a set of well-fitting segmentation fragments. Patch-PerPixMatch is computationally efficient albeit being full 3d, while also highly robust to inaccuracies in the automated neuron instance segmentation. We are releasing PatchPerPixMatch search results for ~30,000 neuron morphologies from the Drosophila hemibrain in ~20,000 MCFO acquisitions of ~3,500 Gen1 GAL4 lines. Code: https://github.com/Kainmueller-Lab/PatchPerPixMatch Results: https://pppm.janelia.org
Open AccessPatchPerPixMatch for Automated 3d Search of Neuronal Morphologies in Light Microscopy
Studies of individual neurons in the Drosophila nervous system are facilitated by transgenic lines that sparsely and repeatably label respective neurons of interest. Sparsity can be enhanced by means of intersectional approaches like the split-GAL4 system, which labels the positive intersection of the expression patterns of two (denser) GAL4 lines. To this end, two GAL4 lines have to be identified as labelling a neuron of interest. Current approaches to tackling this task include visual inspection, as well as automated search in 2d projection images, of single cell multi-color flip-out (MCFO) acquisitions of GAL4 expression patterns. There is to date no automated method available that performs full 3d search in MCFO imagery of GAL4 lines, nor one that leverages automated reconstructions of the labelled neuron morphologies. To close this gap, we propose PatchPerPixMatch, a fully automated approach for finding a given neuron morphology in MCFO acquisitions of Gen1 GAL4 lines. PatchPerPixMatch performs automated instance segmentation of MCFO acquisitions, and subsequently searches for a target neuron morphology by minimizing an objective that aims at covering the target with a set of well-fitting segmentation fragments. Patch-PerPixMatch is computationally efficient albeit being full 3d, while also highly robust to inaccuracies in the automated neuron instance segmentation. We are releasing PatchPerPixMatch search results for ~30,000 neuron morphologies from the Drosophila hemibrain in ~20,000 MCFO acquisitions of ~3,500 Gen1 GAL4 lines. Code: https://github.com/Kainmueller-Lab/PatchPerPixMatch Results: https://pppm.janelia.org
- Top Cited PapersOpen Access
- 19 March 2020
- Published by Springer Nature in Communications Physics
- Vol. 3 (1) , 1-8
Artificial-intelligence-driven scanning probe microscopy
Scanning probe microscopy (SPM) has revolutionized the fields of materials, nano-science, chemistry, and biology, by enabling mapping of surface properties and surface manipulation with atomic precision. However, these achievements require constant human supervision; fully automated SPM has not been accomplished yet. Here we demonstrate an artificial intelligence framework based on machine learning for autonomous SPM operation (DeepSPM). DeepSPM includes an algorithmic search of good sample regions, a convolutional neural network to assess the quality of acquired images, and a deep reinforcement learning agent to reliably condition the state of the probe. DeepSPM is able to acquire and classify data continuously in multi-day scanning tunneling microscopy experiments, managing the probe quality in response to varying experimental conditions. Our approach paves the way for advanced methods hardly feasible by human operation (e.g., large dataset acquisition and SPM-based nanolithography). DeepSPM can be generalized to most SPM techniques, with the source code publicly available.
- 19 March 2020
- Published by Springer Nature in Communications Physics
- Vol. 3 (1) , 1-8
Top Cited PapersOpen AccessArtificial-intelligence-driven scanning probe microscopy
Scanning probe microscopy (SPM) has revolutionized the fields of materials, nano-science, chemistry, and biology, by enabling mapping of surface properties and surface manipulation with atomic precision. However, these achievements require constant human supervision; fully automated SPM has not been accomplished yet. Here we demonstrate an artificial intelligence framework based on machine learning for autonomous SPM operation (DeepSPM). DeepSPM includes an algorithmic search of good sample regions, a convolutional neural network to assess the quality of acquired images, and a deep reinforcement learning agent to reliably condition the state of the probe. DeepSPM is able to acquire and classify data continuously in multi-day scanning tunneling microscopy experiments, managing the probe quality in response to varying experimental conditions. Our approach paves the way for advanced methods hardly feasible by human operation (e.g., large dataset acquisition and SPM-based nanolithography). DeepSPM can be generalized to most SPM techniques, with the source code publicly available.
- Open Access
An Auxiliary Task for Learning Nuclei Segmentation in 3D Microscopy Images
Segmentation of cell nuclei in microscopy images is a prevalent necessity in cell biology. Especially for three-dimensional datasets, manual segmentation is prohibitively time-consuming, motivating the need for automated methods. Learning-based methods trained on pixel-wise ground-truth segmentations have been shown to yield state-of-the-art results on 2d benchmark image data of nuclei, yet a respective benchmark is missing for 3d image data. In this work, we perform a comparative evaluation of nuclei segmentation algorithms on a database of manually segmented 3d light microscopy volumes. We propose a novel learning strategy that boosts segmentation accuracy by means of a simple auxiliary task, thereby robustly outperforming each of our baselines. Furthermore, we show that one of our baselines, the popular three-label model, when trained with our proposed auxiliary task, outperforms the recent StarDist-3D. As an additional, practical contribution, we benchmark nuclei segmentation against nuclei detection, i.e. the task of merely pinpointing individual nuclei without generating respective pixel-accurate segmentations. For learning nuclei detection, large 3d training datasets of manually annotated nuclei center points are available. However, the impact on detection accuracy caused by training on such sparse ground truth as opposed to dense pixel-wise ground truth has not yet been quantified. To this end, we compare nuclei detection accuracy yielded by training on dense vs. sparse ground truth. Our results suggest that training on sparse ground truth yields competitive nuclei detection rates.
Open AccessAn Auxiliary Task for Learning Nuclei Segmentation in 3D Microscopy Images
Segmentation of cell nuclei in microscopy images is a prevalent necessity in cell biology. Especially for three-dimensional datasets, manual segmentation is prohibitively time-consuming, motivating the need for automated methods. Learning-based methods trained on pixel-wise ground-truth segmentations have been shown to yield state-of-the-art results on 2d benchmark image data of nuclei, yet a respective benchmark is missing for 3d image data. In this work, we perform a comparative evaluation of nuclei segmentation algorithms on a database of manually segmented 3d light microscopy volumes. We propose a novel learning strategy that boosts segmentation accuracy by means of a simple auxiliary task, thereby robustly outperforming each of our baselines. Furthermore, we show that one of our baselines, the popular three-label model, when trained with our proposed auxiliary task, outperforms the recent StarDist-3D. As an additional, practical contribution, we benchmark nuclei segmentation against nuclei detection, i.e. the task of merely pinpointing individual nuclei without generating respective pixel-accurate segmentations. For learning nuclei detection, large 3d training datasets of manually annotated nuclei center points are available. However, the impact on detection accuracy caused by training on such sparse ground truth as opposed to dense pixel-wise ground truth has not yet been quantified. To this end, we compare nuclei detection accuracy yielded by training on dense vs. sparse ground truth. Our results suggest that training on sparse ground truth yields competitive nuclei detection rates.
- 1 May 1993
- Published by Elsevier in Materials Science and Engineering: A
- Vol. 164 (1-2) , 135-140
Dislocation dynamics and brittle-to-ductile transitions
- 1 May 1993
- Published by Elsevier in Materials Science and Engineering: A
- Vol. 164 (1-2) , 135-140
Dislocation dynamics and brittle-to-ductile transitions