Publiée 22 juillet 2026
Doctorant F/H PhD in bioimage processing - Convolutional neural network for the segmentation of astrocytes in volume electron microscopy datasets
Inria
Villeurbanne, Auvergne-Rhône-Alpes 69100, France
CDI
A propos du centre ou de la direction fonctionnelle
Le centre Inria de Lyon est le 9ème centre de recherche Inria. Créé en janvier 2022, il regroupe environ 410 personnes au sein de 20 équipes de recherche et des services supports à la recherche.
Ses équipes sont localisées à Villeurbanne, à Lyon Gerland, Lyon Bron ainsi qu'à Saint-Etienne.
Le centre de Lyon est présent dans les domaines du logiciel, du calcul distribué et haute performance, des systèmes embarqués, du calcul quantique et de respect de la vie privée dans le monde numérique, mais aussi de la santé et de la biologie numériques.
Contexte et atouts du poste
This PhD project is part of a collaboration between two Inria teams, SAIRPICO and AIstroSight, and two teams from
McGill University, Montréal, Canada: the Shape Analysis lab, led by Pr. K. Siddiqi, and the Murai lab.
Astrocytes are glial cells of the central nervous system involved in numerous brain functions, such as the regulation of neurotransmission, synaptogenesis, as well as the maintenance of ionic and metabolic homeostasis [1]. The study of astrocyte morphology is a growing field of research. Alterations in their shape may compromise their ability to effectively support brain functions, thus contributing to the progression of neurodegenerative processes [2]. Understanding the shape and connectivity of astrocytes as well as their variability depending on the cerebral and pathological context is therefore essential to deepen our knowledge of the mechanisms involved in brain diseases.
Electron microscopy (EM) provides the highest resolution of cell structure. Recent technical advances such as cryo- and volume EM (vEM) have yielded isotropic reconstructions of cells and organelles at an unprecedented spatial resolution (< 10 nm3) [3] . High-resolution vEM data have revealed that these cells exhibit extremely complex morphologies, connected to each other through nanometric branches 2 (Figure 1). This complexity makes their segmentation extremely difficult.
While state-of-the-art image segmentation methods have been effective for cells with more regular shapes such as neurons, the segmentation of astrocytes remains a major challenge. To the best of our knowledge, no segmentation method has yet been specifically tailored to astrocytes, primarily due to the complexity of their morphology [4].
Mission confiée
Our goal is to characterize the diversity of the structure and connectivity of astrocytes and to investigate how it affects brain function. To do so, high-quality simulation-ready 3D meshes of astrocytes are needed. However, no tool is currently available to automatically segment astrocytes from vEM datasets, so it takes years for trained experts to obtain annotations and 3D reconstructions.
The Allen Institute's MICrONS Explorer cortical MM3 dataset is the largest EM dataset of a mammalian brain circuit to date (2 PB of about 4 nm resolution images of a 1.4mm x .87mm x .84 mm volume), comprising 200,000 cells from multiple functional areas [5,6]. In this dataset, astrocytes were automatically segmented using tools trained with neuron annotations, resulting in numerous errors (Figure 2) [7]. To overcome this limitation, we aim to leverage convolutional neural networks (CNNs) to develop a method for segmenting astrocytes. Published [8-11] and unpublished annotated ground-truth data will be used to train and validate the model. One avenue we would like to explore is the integration of prior knowledge from geometry and shape analysis into the network, together with probabilistic information concerning other annotated structures in the images.
One of the main challenges will be to generalize this method to datasets in which astrocytes were not segmented or acquired under different experimental conditions (microscope, fixation protocol, spatial resolution, pathological context, subject age), which leads to variations in astrocyte morphology and in the signal-to-noise ratio of the images. Then, 3D astrocyte models will be generated and the resulting models and meshes will populate an open-access database currently being developed by the AIstroSight team.
Overall, this project will contribute to enriching our understanding of the basic organizing principles and the variability of astrocyte structure in physiological (species, brain region, developmental stage, aging) and pathophysiological conditions. Moreover, computational modeling will allow us to predict how these structural properties impact astrocyte function. As astrocytes contact hundreds of thousands of synapses simultaneously, this study will provide novel insights into computational principles of information integration in the brain.
References
1. Verkhratsky, A. & Nedergaard, M. Physiology of Astroglia. Physiol. Rev. 98, 239-389 (2018).
2. Baldwin, K. T., Murai, K. K. & Khakh, B. S. Astrocyte morphology. Trends in Cell Biology https://doi.org/10.1016/j.tcb.2023.09.006 (2023) doi:10.1016/j.tcb.2023.09.006.
3. Peddie, C. J. et al. Volume electron microscopy. Nat Rev Methods Primers 2, 1-23 (2022).
4. Baldwin, K. T., Murai, K. K. & Khakh, B. S. Astrocyte morphology. Trends in Cell Biology https://doi.org/10.1016/j.tcb.2023.09.006 (2023) doi:10.1016/j.tcb.2023.09.006.
5. Consortium, T. Mic. et al. Functional connectomics spanning multiple areas of mouse visual cortex. 2021.07.28.454025 Preprint at https://doi.org/10.1101/2021.07.28.454025 (2023).
6. Bae, J. A. et al. Functional connectomics spanning multiple areas of mouse visual cortex. Nature 640, 435-447 (2025).
7. Syed, T. A. et al. Beyond neurons: computer vision methods for analysis of morphologically complex astrocytes. Front. Comput. Sci. 6, (2024).
8. Salmon, C. K. et al. Organizing principles of astrocytic nanoarchitecture in the mouse cerebral cortex. Current Biology 0, (2023).
9. Aten, S. et al. Ultrastructural view of astrocyte arborization, astrocyte-astrocyte and astrocyte-synapse contacts, intracellular vesicle-like structures, and mitochondrial network. Progress in Neurobiology 213, 102264 (2022).
10. Cali, C. et al. 3D cellular reconstruction of cortical glia and parenchymal morphometric analysis from Serial Block-Face Electron Microscopy of juvenile rat. Progress in Neurobiology 101696 (2019) doi:10.1016/j.pneurobio.2019.101696.
11. Benoit, L. et al. Astrocytes functionally integrate multiple synapses via specialized leaflet domains. Cell https://doi.org/10.1016/j.cell.2025.08.036 (2025) doi:10.1016/j.cell.2025.08.036.
Principales activités
In this context, the candidate is expected to:
- conduct a literature review on astrocyte segmentation approaches, particularly on EM data;
- conduct a literature review on the integration of prior knowledge from shape analysis into a network;
- develop a CNN-based segmentation method, building on the literature review;
- experimentally evaluate the algorithm on annotated EM data and quantitatively compare it to existing segmentation methods;
- investigate the generalization capability of the developed method on other EM datasets with different characteristics (signal-to-noise ratio, spatial resolution, etc.);
- study how astrocyte structural properties and connectivity affect information processing in the brain
Compétences
The candidate should preferably hold a Master's degree in computer science and demonstrate skills or experience in most of the following areas:
- Image processing and analysis
- Machine learning
- Deep learning (CNNs)
- Excellent programming skills in Python
- Strong interest in biological applications
- Excellent command of English (oral and written)
- Aptitude for working in a team and developing external collaborations
Additional qualifications that will be valued include:
- Strong autonomy and demonstrated ability to take initiative
- Sound, well-justified technical decision-making (trade-offs, documentation, rationale)
- Knowledge of cell biology
- Familiarity with electron microscopy datasets
- Knowledge in shape analysis
Avantages
Rémunération
2300 € brut mensuel
Le centre Inria de Lyon est le 9ème centre de recherche Inria. Créé en janvier 2022, il regroupe environ 410 personnes au sein de 20 équipes de recherche et des services supports à la recherche.
Ses équipes sont localisées à Villeurbanne, à Lyon Gerland, Lyon Bron ainsi qu'à Saint-Etienne.
Le centre de Lyon est présent dans les domaines du logiciel, du calcul distribué et haute performance, des systèmes embarqués, du calcul quantique et de respect de la vie privée dans le monde numérique, mais aussi de la santé et de la biologie numériques.
Contexte et atouts du poste
This PhD project is part of a collaboration between two Inria teams, SAIRPICO and AIstroSight, and two teams from
McGill University, Montréal, Canada: the Shape Analysis lab, led by Pr. K. Siddiqi, and the Murai lab.
Astrocytes are glial cells of the central nervous system involved in numerous brain functions, such as the regulation of neurotransmission, synaptogenesis, as well as the maintenance of ionic and metabolic homeostasis [1]. The study of astrocyte morphology is a growing field of research. Alterations in their shape may compromise their ability to effectively support brain functions, thus contributing to the progression of neurodegenerative processes [2]. Understanding the shape and connectivity of astrocytes as well as their variability depending on the cerebral and pathological context is therefore essential to deepen our knowledge of the mechanisms involved in brain diseases.
Electron microscopy (EM) provides the highest resolution of cell structure. Recent technical advances such as cryo- and volume EM (vEM) have yielded isotropic reconstructions of cells and organelles at an unprecedented spatial resolution (< 10 nm3) [3] . High-resolution vEM data have revealed that these cells exhibit extremely complex morphologies, connected to each other through nanometric branches 2 (Figure 1). This complexity makes their segmentation extremely difficult.
While state-of-the-art image segmentation methods have been effective for cells with more regular shapes such as neurons, the segmentation of astrocytes remains a major challenge. To the best of our knowledge, no segmentation method has yet been specifically tailored to astrocytes, primarily due to the complexity of their morphology [4].
Mission confiée
Our goal is to characterize the diversity of the structure and connectivity of astrocytes and to investigate how it affects brain function. To do so, high-quality simulation-ready 3D meshes of astrocytes are needed. However, no tool is currently available to automatically segment astrocytes from vEM datasets, so it takes years for trained experts to obtain annotations and 3D reconstructions.
The Allen Institute's MICrONS Explorer cortical MM3 dataset is the largest EM dataset of a mammalian brain circuit to date (2 PB of about 4 nm resolution images of a 1.4mm x .87mm x .84 mm volume), comprising 200,000 cells from multiple functional areas [5,6]. In this dataset, astrocytes were automatically segmented using tools trained with neuron annotations, resulting in numerous errors (Figure 2) [7]. To overcome this limitation, we aim to leverage convolutional neural networks (CNNs) to develop a method for segmenting astrocytes. Published [8-11] and unpublished annotated ground-truth data will be used to train and validate the model. One avenue we would like to explore is the integration of prior knowledge from geometry and shape analysis into the network, together with probabilistic information concerning other annotated structures in the images.
One of the main challenges will be to generalize this method to datasets in which astrocytes were not segmented or acquired under different experimental conditions (microscope, fixation protocol, spatial resolution, pathological context, subject age), which leads to variations in astrocyte morphology and in the signal-to-noise ratio of the images. Then, 3D astrocyte models will be generated and the resulting models and meshes will populate an open-access database currently being developed by the AIstroSight team.
Overall, this project will contribute to enriching our understanding of the basic organizing principles and the variability of astrocyte structure in physiological (species, brain region, developmental stage, aging) and pathophysiological conditions. Moreover, computational modeling will allow us to predict how these structural properties impact astrocyte function. As astrocytes contact hundreds of thousands of synapses simultaneously, this study will provide novel insights into computational principles of information integration in the brain.
References
1. Verkhratsky, A. & Nedergaard, M. Physiology of Astroglia. Physiol. Rev. 98, 239-389 (2018).
2. Baldwin, K. T., Murai, K. K. & Khakh, B. S. Astrocyte morphology. Trends in Cell Biology https://doi.org/10.1016/j.tcb.2023.09.006 (2023) doi:10.1016/j.tcb.2023.09.006.
3. Peddie, C. J. et al. Volume electron microscopy. Nat Rev Methods Primers 2, 1-23 (2022).
4. Baldwin, K. T., Murai, K. K. & Khakh, B. S. Astrocyte morphology. Trends in Cell Biology https://doi.org/10.1016/j.tcb.2023.09.006 (2023) doi:10.1016/j.tcb.2023.09.006.
5. Consortium, T. Mic. et al. Functional connectomics spanning multiple areas of mouse visual cortex. 2021.07.28.454025 Preprint at https://doi.org/10.1101/2021.07.28.454025 (2023).
6. Bae, J. A. et al. Functional connectomics spanning multiple areas of mouse visual cortex. Nature 640, 435-447 (2025).
7. Syed, T. A. et al. Beyond neurons: computer vision methods for analysis of morphologically complex astrocytes. Front. Comput. Sci. 6, (2024).
8. Salmon, C. K. et al. Organizing principles of astrocytic nanoarchitecture in the mouse cerebral cortex. Current Biology 0, (2023).
9. Aten, S. et al. Ultrastructural view of astrocyte arborization, astrocyte-astrocyte and astrocyte-synapse contacts, intracellular vesicle-like structures, and mitochondrial network. Progress in Neurobiology 213, 102264 (2022).
10. Cali, C. et al. 3D cellular reconstruction of cortical glia and parenchymal morphometric analysis from Serial Block-Face Electron Microscopy of juvenile rat. Progress in Neurobiology 101696 (2019) doi:10.1016/j.pneurobio.2019.101696.
11. Benoit, L. et al. Astrocytes functionally integrate multiple synapses via specialized leaflet domains. Cell https://doi.org/10.1016/j.cell.2025.08.036 (2025) doi:10.1016/j.cell.2025.08.036.
Principales activités
In this context, the candidate is expected to:
- conduct a literature review on astrocyte segmentation approaches, particularly on EM data;
- conduct a literature review on the integration of prior knowledge from shape analysis into a network;
- develop a CNN-based segmentation method, building on the literature review;
- experimentally evaluate the algorithm on annotated EM data and quantitatively compare it to existing segmentation methods;
- investigate the generalization capability of the developed method on other EM datasets with different characteristics (signal-to-noise ratio, spatial resolution, etc.);
- study how astrocyte structural properties and connectivity affect information processing in the brain
Compétences
The candidate should preferably hold a Master's degree in computer science and demonstrate skills or experience in most of the following areas:
- Image processing and analysis
- Machine learning
- Deep learning (CNNs)
- Excellent programming skills in Python
- Strong interest in biological applications
- Excellent command of English (oral and written)
- Aptitude for working in a team and developing external collaborations
Additional qualifications that will be valued include:
- Strong autonomy and demonstrated ability to take initiative
- Sound, well-justified technical decision-making (trade-offs, documentation, rationale)
- Knowledge of cell biology
- Familiarity with electron microscopy datasets
- Knowledge in shape analysis
Avantages
- Restauration subventionnée
- Transports publics remboursés partiellement
- Congés: 7 semaines de congés annuels + 10 jours de RTT (base temps plein) + possibilité d'autorisations d'absence exceptionnelle (ex : enfants malades, déménagement)
- Possibilité de télétravail (après 6 mois d'ancienneté) et aménagement du temps de travail
- Équipements professionnels à disposition (visioconférence, prêts de matériels informatiques, etc.)
- Prestations sociales, culturelles et sportives (Association de gestion des œuvres sociales d'Inria)
- Accès à la formation professionnelle
- Sécurité sociale
Rémunération
2300 € brut mensuel
