Hamza Kebiri
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Hamza Kebiri has a total of 89 co-authors. Here are the top co-authors by number of shared publications:
- Meritxell Bach Cuadra
CIBM Center for Biomedical Imaging, Lausanne, Switzerland
16 shared publications - Ali Gholipour
Harvard Medical School
6 shared publications - Davood Karimi
Computational Radiology Laboratory, Boston Children’s Hospital, Boston, MA, USA
5 shared publications - Hélène Lajous
Department of Radiology, Lausanne University Hospital (CHUV), Lausanne, Switzerland
5 shared publications - Rizhong Lin
Signal Processing Laboratory (LTS5) Ecole Polytechnique Fédérale de Lausanne (EPFL) Lausanne Switzerland
5 shared publications - Jean-Philippe Thiran
LTS5, Ecole Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland
4 shared publications
Latest Publications
- Open Access
- 3 September 2026
- Published by Springer Nature in npj Science of Learning
Learning environments are associated with age-related differences in thalamocortical attention circuits in childhood
The thalamus plays a central role in sensory gating and large-scale cortical communication, but its developmental organization in relation to real-world learning environments has rarely been examined. Here, we analyzed resting-state fMRI data from 72 children aged 4–15 years enrolled in Montessori or traditional classrooms. Using a data-driven parcellation approach based on voxelwise functional connectivity, we identified five spatially coherent thalamic territories with distinct large-scale connectivity profiles, primarily linked to visual, sensorimotor, and attention networks. In a thalamic territory preferentially coupled to the ventral attention system, Montessori enrollment was associated with stronger thalamocortical connectivity and a stronger positive association between age and FC. No clusters showed enhanced connectivity with executive-control or default-mode networks, consistent with their later maturation. Together, these findings suggest that schooling context co-varies with the developmental refinement of thalamic circuits supporting salience detection and the selective prioritization of information during learning. Because the design is cross-sectional and families self-select into schools, causal direction cannot be inferred and unmeasured family-level factors (e.g., parental values, home learning environment) may contribute to these associations. Nonetheless, the results identify a tractable neural pathway through which everyday learning environments may relate to developing brain systems underlying attention and adaptive behaviour.
- 3 September 2026
- Published by Springer Nature in npj Science of Learning
Open AccessLearning environments are associated with age-related differences in thalamocortical attention circuits in childhood
The thalamus plays a central role in sensory gating and large-scale cortical communication, but its developmental organization in relation to real-world learning environments has rarely been examined. Here, we analyzed resting-state fMRI data from 72 children aged 4–15 years enrolled in Montessori or traditional classrooms. Using a data-driven parcellation approach based on voxelwise functional connectivity, we identified five spatially coherent thalamic territories with distinct large-scale connectivity profiles, primarily linked to visual, sensorimotor, and attention networks. In a thalamic territory preferentially coupled to the ventral attention system, Montessori enrollment was associated with stronger thalamocortical connectivity and a stronger positive association between age and FC. No clusters showed enhanced connectivity with executive-control or default-mode networks, consistent with their later maturation. Together, these findings suggest that schooling context co-varies with the developmental refinement of thalamic circuits supporting salience detection and the selective prioritization of information during learning. Because the design is cross-sectional and families self-select into schools, causal direction cannot be inferred and unmeasured family-level factors (e.g., parental values, home learning environment) may contribute to these associations. Nonetheless, the results identify a tractable neural pathway through which everyday learning environments may relate to developing brain systems underlying attention and adaptive behaviour.
- 1 July 2026
- Published by Machine Learning for Biomedical Imaging in Machine Learning for Biomedical Imaging
- Vol. 2026 (July 2026) , 482-506
Evaluating Synthetic Data Generation for Domain Generalization in Fetal Brain MRI Segmentation
Fetal brain tissue segmentation from magnetic resonance imaging (MRI) is crucial for studying neurodevelopment, but remains challenging due to data heterogeneity and limited annotations. Domain randomization (DR) has recently emerged as a promising strategy for single-source domain generalization by synthesizing training images with randomized artifacts, contrast, and resolution. In this work, we investigate how to maximize the out-of-domain (OOD) generalization of DR-based methods using FetalSynthSeg as a case study. We show that simple Gaussian mixture-based intensity modeling outperforms more complex physics-based simulations and that intensity clustering (subdividing tissue classes by intensity) substantially improves OOD robustness. Evaluated on 348 fetal subjects from four sites spanning 0.55–3T and both T1w and T2w contrasts, FetalSynthSeg reaches state-of-the-art performance on several FeTA 2024 testing datasets (80–85 Dice score) and, for the first time, offers robust segmentation on modalities other than T2w for fetal brain segmentation (80 Dice on dHCP-T1w dataset). Compared with state-of-the-art methods such as BOUNTI, nnU-Net ensemble, and the FeTA 2024 winner, FetalSynthSeg delivers comparable or superior accuracy while maintaining strong robustness across domain shifts. Our code, model weights, and Docker image ready for easy inference are available at https://hub.docker.com/r/vzalevskyi/fetalsynthseg
- 1 July 2026
- Published by Machine Learning for Biomedical Imaging in Machine Learning for Biomedical Imaging
- Vol. 2026 (July 2026) , 482-506
Evaluating Synthetic Data Generation for Domain Generalization in Fetal Brain MRI Segmentation
Fetal brain tissue segmentation from magnetic resonance imaging (MRI) is crucial for studying neurodevelopment, but remains challenging due to data heterogeneity and limited annotations. Domain randomization (DR) has recently emerged as a promising strategy for single-source domain generalization by synthesizing training images with randomized artifacts, contrast, and resolution. In this work, we investigate how to maximize the out-of-domain (OOD) generalization of DR-based methods using FetalSynthSeg as a case study. We show that simple Gaussian mixture-based intensity modeling outperforms more complex physics-based simulations and that intensity clustering (subdividing tissue classes by intensity) substantially improves OOD robustness. Evaluated on 348 fetal subjects from four sites spanning 0.55–3T and both T1w and T2w contrasts, FetalSynthSeg reaches state-of-the-art performance on several FeTA 2024 testing datasets (80–85 Dice score) and, for the first time, offers robust segmentation on modalities other than T2w for fetal brain segmentation (80 Dice on dHCP-T1w dataset). Compared with state-of-the-art methods such as BOUNTI, nnU-Net ensemble, and the FeTA 2024 winner, FetalSynthSeg delivers comparable or superior accuracy while maintaining strong robustness across domain shifts. Our code, model weights, and Docker image ready for easy inference are available at https://hub.docker.com/r/vzalevskyi/fetalsynthseg
- Open Access
- 7 October 2025
- Published by Wiley in Human Brain Mapping
- Vol. 46 (14) , e70367
Deep Learning for fODF Estimation in Infant Brains: Model Comparison, Ground‐Truth Impact, and Domain Shift Mitigation
The accurate estimation of fiber orientation distribution functions (fODFs) in diffusion magnetic resonance imaging (MRI) is crucial for understanding early brain development and its potential disruptions. Although supervised deep learning (DL) models have shown promise in fODF estimation from neonatal diffusion MRI (dMRI) data, the out‐of‐domain (OOD) performance of these models remains largely unexplored, especially under diverse domain shift scenarios. This study evaluated the robustness of three state‐of‐the‐art DL architectures: multilayer perceptron (MLP), transformer, and U‐Net/convolutional neural network (CNN) on fODF predictions derived from dMRI data. Using 488 subjects from the developing Human Connectome Project (dHCP) and the Baby Connectome Project (BCP) datasets, we reconstructed reference fODFs from the full dMRI series using single‐shell three‐tissue constrained spherical deconvolution (SS3T‐CSD) and multi‐shell multi‐tissue CSD (MSMT‐CSD) to generate reference fODF reconstructions for model training, and systematically assessed the impact of age, scanner/protocol differences, and input dimensionality on model performance. Our findings reveal that U‐Net consistently outperformed other models when fewer diffusion gradient directions were used, particularly with the SS3T‐CSD‐derived ground truth, which showed superior performance in capturing crossing fibers. However, as the number of input diffusion gradient directions increased, MLP and the transformer‐based model exhibited steady gains in accuracy. Nevertheless, performance nearly plateaued from 28 to 45 input directions in all models. Age‐related domain shifts showed asymmetric patterns, being less pronounced in late developmental stages (late neonates, and babies), with SS3T‐CSD demonstrating greater robustness to variability compared to MSMT‐CSD. To address inter‐site domain shifts, we implemented two adaptation strategies: the Method of Moments (MoM) and fine‐tuning. Both strategies achieved significant improvements () in over 95% of tested configurations, with fine‐tuning consistently yielding superior results and U‐Net benefiting the most from increased target subjects. This study represents the first systematic evaluation of OOD settings in DL applications to fODF estimation, providing critical insights into model robustness and adaptation strategies for diverse clinical and research applications.
- 7 October 2025
- Published by Wiley in Human Brain Mapping
- Vol. 46 (14) , e70367
Open AccessDeep Learning for fODF Estimation in Infant Brains: Model Comparison, Ground‐Truth Impact, and Domain Shift Mitigation
The accurate estimation of fiber orientation distribution functions (fODFs) in diffusion magnetic resonance imaging (MRI) is crucial for understanding early brain development and its potential disruptions. Although supervised deep learning (DL) models have shown promise in fODF estimation from neonatal diffusion MRI (dMRI) data, the out‐of‐domain (OOD) performance of these models remains largely unexplored, especially under diverse domain shift scenarios. This study evaluated the robustness of three state‐of‐the‐art DL architectures: multilayer perceptron (MLP), transformer, and U‐Net/convolutional neural network (CNN) on fODF predictions derived from dMRI data. Using 488 subjects from the developing Human Connectome Project (dHCP) and the Baby Connectome Project (BCP) datasets, we reconstructed reference fODFs from the full dMRI series using single‐shell three‐tissue constrained spherical deconvolution (SS3T‐CSD) and multi‐shell multi‐tissue CSD (MSMT‐CSD) to generate reference fODF reconstructions for model training, and systematically assessed the impact of age, scanner/protocol differences, and input dimensionality on model performance. Our findings reveal that U‐Net consistently outperformed other models when fewer diffusion gradient directions were used, particularly with the SS3T‐CSD‐derived ground truth, which showed superior performance in capturing crossing fibers. However, as the number of input diffusion gradient directions increased, MLP and the transformer‐based model exhibited steady gains in accuracy. Nevertheless, performance nearly plateaued from 28 to 45 input directions in all models. Age‐related domain shifts showed asymmetric patterns, being less pronounced in late developmental stages (late neonates, and babies), with SS3T‐CSD demonstrating greater robustness to variability compared to MSMT‐CSD. To address inter‐site domain shifts, we implemented two adaptation strategies: the Method of Moments (MoM) and fine‐tuning. Both strategies achieved significant improvements () in over 95% of tested configurations, with fine‐tuning consistently yielding superior results and U‐Net benefiting the most from increased target subjects. This study represents the first systematic evaluation of OOD settings in DL applications to fODF estimation, providing critical insights into model robustness and adaptation strategies for diverse clinical and research applications.
- Open Access
- 24 July 2025
- Published by MIT Press in Imaging Neuroscience
- Vol. 3
Functional organization of the neonatal thalamus across development depicted by functional MRI
The thalamus is a central component of the brain that is involved in a variety of functions, from sensory processing to high-order cognition. Its structure and function in the first weeks of extrauterine life, including its connections to different cortical and subcortical areas, have not yet been widely explored. Here, we used functional magnetic resonance imaging data of 588 newborns during natural sleep from the developing Human Connectome Project to study the functional organization of the thalamus from 37 to 44 post-conceptual weeks. We introduce KNIT: K-means for Nuclei in Infant Thalamus. The framework employs a highly granular vector space of 40 features, each corresponding to functional connectivity to a brain region, using k-means clustering and uncertainty quantification through bootstrapping to delineate thalamic units. Although the different clusters showed common patterns of increased connectivity to the superior temporal gyrus, the parietal, and the frontal cortex, implying an expected decrease in specialization at that age, they also show some specificity. That is, a pulvinar unit was identified, similar to the adult thalamus. Ventrolateral motor and medial salience units were also highlighted. The latter appeared around 41 weeks of age, while the former showed at least from 37 weeks, but had a decrease in relative volume through age, replaced mostly by a dominant dorsal thalamic unit. We also observed an increase in clustering robustness and in hemispheric bilateral symmetry with age, suggesting more specialized functional units. We also found a burst in global thalamic connectivity around 41 weeks. Finally, we demonstrate the benefits of this method in terms of granularity compared to the more conventional winner-takes-all approach.
- 24 July 2025
- Published by MIT Press in Imaging Neuroscience
- Vol. 3
Open AccessFunctional organization of the neonatal thalamus across development depicted by functional MRI
The thalamus is a central component of the brain that is involved in a variety of functions, from sensory processing to high-order cognition. Its structure and function in the first weeks of extrauterine life, including its connections to different cortical and subcortical areas, have not yet been widely explored. Here, we used functional magnetic resonance imaging data of 588 newborns during natural sleep from the developing Human Connectome Project to study the functional organization of the thalamus from 37 to 44 post-conceptual weeks. We introduce KNIT: K-means for Nuclei in Infant Thalamus. The framework employs a highly granular vector space of 40 features, each corresponding to functional connectivity to a brain region, using k-means clustering and uncertainty quantification through bootstrapping to delineate thalamic units. Although the different clusters showed common patterns of increased connectivity to the superior temporal gyrus, the parietal, and the frontal cortex, implying an expected decrease in specialization at that age, they also show some specificity. That is, a pulvinar unit was identified, similar to the adult thalamus. Ventrolateral motor and medial salience units were also highlighted. The latter appeared around 41 weeks of age, while the former showed at least from 37 weeks, but had a decrease in relative volume through age, replaced mostly by a dominant dorsal thalamic unit. We also observed an increase in clustering robustness and in hemispheric bilateral symmetry with age, suggesting more specialized functional units. We also found a burst in global thalamic connectivity around 41 weeks. Finally, we demonstrate the benefits of this method in terms of granularity compared to the more conventional winner-takes-all approach.
- Open Access
- 18 April 2025
- Published by Springer Nature
- p. 24-34
Ground-Truth Effects in Learning-Based Fiber Orientation Distribution Estimation in Neonatal Brains
Diffusion Magnetic Resonance Imaging (dMRI) is a non-invasive method for depicting brain microstructure in vivo. Fiber orientation distributions (FODs) are mathematical representations extensively used to map white matter fiber configurations. Recently, FOD estimation with deep neural networks has seen growing success, in particular, those of neonates estimated with fewer diffusion measurements. These methods are mostly trained on target FODs reconstructed with multi-shell multi-tissue constrained spherical deconvolution (MSMT-CSD), which might not be the ideal ground truth for developing brains. Here, we investigate this hypothesis by training a state-of-the-art model based on the U-Net architecture on both MSMT-CSD and single-shell three-tissue constrained spherical deconvolution (SS3T-CSD). Our results suggest that SS3T-CSD might be more suited for neonatal brains, given that the ratio between single and multiple fiber-estimated voxels with SS3T-CSD is more realistic compared to MSMT-CSD. Additionally, increasing the number of input gradient directions significantly improves performance with SS3T-CSD over MSMT-CSD. Finally, in an age domain-shift setting, SS3T-CSD maintains robust performance across age groups, indicating its potential for more accurate neonatal brain imaging.
- 18 April 2025
- Published by Springer Nature
- p. 24-34
Open AccessGround-Truth Effects in Learning-Based Fiber Orientation Distribution Estimation in Neonatal Brains
Diffusion Magnetic Resonance Imaging (dMRI) is a non-invasive method for depicting brain microstructure in vivo. Fiber orientation distributions (FODs) are mathematical representations extensively used to map white matter fiber configurations. Recently, FOD estimation with deep neural networks has seen growing success, in particular, those of neonates estimated with fewer diffusion measurements. These methods are mostly trained on target FODs reconstructed with multi-shell multi-tissue constrained spherical deconvolution (MSMT-CSD), which might not be the ideal ground truth for developing brains. Here, we investigate this hypothesis by training a state-of-the-art model based on the U-Net architecture on both MSMT-CSD and single-shell three-tissue constrained spherical deconvolution (SS3T-CSD). Our results suggest that SS3T-CSD might be more suited for neonatal brains, given that the ratio between single and multiple fiber-estimated voxels with SS3T-CSD is more realistic compared to MSMT-CSD. Additionally, increasing the number of input gradient directions significantly improves performance with SS3T-CSD over MSMT-CSD. Finally, in an age domain-shift setting, SS3T-CSD maintains robust performance across age groups, indicating its potential for more accurate neonatal brain imaging.
- Open Access
Domain shifts in infant brain diffusion MRI: Quantification and mitigation for fODF estimation
The accurate estimation of fiber orientation distributions (fODFs) in diffusion magnetic resonance imaging (MRI) is crucial for understanding early brain development and its potential disruptions. While deep learning (DL) models have shown promise in fODF estimation from neonatal diffusion MRI (dMRI) data, the out-of-domain (OOD) performance of these models remains largely unexplored, especially under diverse domain shift scenarios. This study evaluated the robustness of three state-of-the-art DL architectures—MLP-based, Transformer-based, and U-Net/CNN-based—on fODF predictions derived from diffusion MRI data. Using 488 subjects from the developing Human Connectome Project (dHCP) and the Baby Connectome Project (BCP) datasets, we reconstructed reference fODFs from the full dMRI series using single-shell three-tissue constrained spherical deconvolution (SS3T-CSD) and multi-shell multi-tissue CSD (MSMT-CSD) to generate reference fODF reconstructions for model training, and systematically assessed the impact of age, scanner/protocol differences, and input dimensionality on model performance. Our findings reveal that U-Net consistently outperformed other models when fewer diffusion gradient directions were used, particularly with the SS3T-CSD-derived ground truth, which showed superior performance in capturing crossing fibers. However, as the number of input diffusion gradient directions increased, MLP and the transformer-based model exhibited steady gains in accuracy. Nevertheless, performance nearly plateaued from 28 to 45 input directions in all models. Age-related domain shifts were found to be less pronounced in late developmental stages (late neonates, and babies), with SS3T-CSD demonstrating greater robustness to variability compared to MSMT-CSD. To address inter-site domain shifts, we implemented two adaptation strategies: the Method of Moments (MoM) and fine-tuning. Fine-tuning consistently yielded superior results, with U-Net benefiting the most from increased target subjects. This study represents the first systematic evaluation of OOD settings in deep learning applications to fODF estimation, providing critical insights into model robustness and adaptation strategies for diverse clinical and research applications.
Open AccessDomain shifts in infant brain diffusion MRI: Quantification and mitigation for fODF estimation
The accurate estimation of fiber orientation distributions (fODFs) in diffusion magnetic resonance imaging (MRI) is crucial for understanding early brain development and its potential disruptions. While deep learning (DL) models have shown promise in fODF estimation from neonatal diffusion MRI (dMRI) data, the out-of-domain (OOD) performance of these models remains largely unexplored, especially under diverse domain shift scenarios. This study evaluated the robustness of three state-of-the-art DL architectures—MLP-based, Transformer-based, and U-Net/CNN-based—on fODF predictions derived from diffusion MRI data. Using 488 subjects from the developing Human Connectome Project (dHCP) and the Baby Connectome Project (BCP) datasets, we reconstructed reference fODFs from the full dMRI series using single-shell three-tissue constrained spherical deconvolution (SS3T-CSD) and multi-shell multi-tissue CSD (MSMT-CSD) to generate reference fODF reconstructions for model training, and systematically assessed the impact of age, scanner/protocol differences, and input dimensionality on model performance. Our findings reveal that U-Net consistently outperformed other models when fewer diffusion gradient directions were used, particularly with the SS3T-CSD-derived ground truth, which showed superior performance in capturing crossing fibers. However, as the number of input diffusion gradient directions increased, MLP and the transformer-based model exhibited steady gains in accuracy. Nevertheless, performance nearly plateaued from 28 to 45 input directions in all models. Age-related domain shifts were found to be less pronounced in late developmental stages (late neonates, and babies), with SS3T-CSD demonstrating greater robustness to variability compared to MSMT-CSD. To address inter-site domain shifts, we implemented two adaptation strategies: the Method of Moments (MoM) and fine-tuning. Fine-tuning consistently yielded superior results, with U-Net benefiting the most from increased target subjects. This study represents the first systematic evaluation of OOD settings in deep learning applications to fODF estimation, providing critical insights into model robustness and adaptation strategies for diverse clinical and research applications.
- Open Access
Functional organization of the neonatal thalamus across development depicted by functional MRI
The thalamus is a central component of the brain that is involved in a variety of functions, from sensory processing to high-order cognition. Its structure and function in the first weeks of extrauterine life, including its connections to different cortical and subcortical areas, have not yet been widely explored. Here, we used resting state functional magnetic resonance imaging data of 730 newborns from the developing Human Connectome Project to study the functional organization of the thalamus from 37 to 44 post-conceptual weeks. We introduce KNIT: K-means for Nuclei in Infant Thalamus. The framework employs a highly granular vector space of 40 features, each corresponding to functional connectivity to a brain region, using k-means clustering and uncertainty quantification through bootstrapping to delineate thalamic units. Although the different clusters showed common patterns of increased connectivity to the superior temporal gyrus, the parietal, and the frontal cortex, implying an expected decrease in specialization at that age, they also show some specificity. That is, a pulvinar unit was identified, similar to the adult thalamus. Ventrolateral motor and medial salience units were also highlighted. The latter appeared around 41 weeks of age, while the former showed at least from 37 weeks, but had a decrease in volume through age, replaced mostly by a dominant dorsal thalamic unit. We also observed an increase in clustering robustness and in hemispheric bilateral symmetry with age, suggesting more specialized functional units. We also found a burst in global thalamic connectivity around 41 weeks. Finally, we demonstrate the benefits of this method in terms of granularity compared to the more conventional winner-takes-all approach.
Open AccessFunctional organization of the neonatal thalamus across development depicted by functional MRI
The thalamus is a central component of the brain that is involved in a variety of functions, from sensory processing to high-order cognition. Its structure and function in the first weeks of extrauterine life, including its connections to different cortical and subcortical areas, have not yet been widely explored. Here, we used resting state functional magnetic resonance imaging data of 730 newborns from the developing Human Connectome Project to study the functional organization of the thalamus from 37 to 44 post-conceptual weeks. We introduce KNIT: K-means for Nuclei in Infant Thalamus. The framework employs a highly granular vector space of 40 features, each corresponding to functional connectivity to a brain region, using k-means clustering and uncertainty quantification through bootstrapping to delineate thalamic units. Although the different clusters showed common patterns of increased connectivity to the superior temporal gyrus, the parietal, and the frontal cortex, implying an expected decrease in specialization at that age, they also show some specificity. That is, a pulvinar unit was identified, similar to the adult thalamus. Ventrolateral motor and medial salience units were also highlighted. The latter appeared around 41 weeks of age, while the former showed at least from 37 weeks, but had a decrease in volume through age, replaced mostly by a dominant dorsal thalamic unit. We also observed an increase in clustering robustness and in hemispheric bilateral symmetry with age, suggesting more specialized functional units. We also found a burst in global thalamic connectivity around 41 weeks. Finally, we demonstrate the benefits of this method in terms of granularity compared to the more conventional winner-takes-all approach.
- Open Access
Ground-truth effects in learning-based fiber orientation distribution estimation in neonatal brains
Diffusion Magnetic Resonance Imaging (dMRI) is a non-invasive method for depicting brain microstructure in vivo. Fiber orientation distributions (FODs) are mathematical representations extensively used to map white matter fiber configurations. Recently, FOD estimation with deep neural networks has seen growing success, in particular, those of neonates estimated with fewer diffusion measurements. These methods are mostly trained on target FODs reconstructed with multi-shell multi-tissue constrained spherical deconvolution (MSMT-CSD), which might not be the ideal ground truth for developing brains. Here, we investigate this hypothesis by training a state-of-the-art model based on the U-Net architecture on both MSMT-CSD and single-shell three-tissue constrained spherical deconvolution (SS3T-CSD). Our results suggest that SS3T-CSD might be more suited for neonatal brains, given that the ratio between single and multiple fiber-estimated voxels with SS3T-CSD is more realistic compared to MSMT-CSD. Additionally, increasing the number of input gradient directions significantly improves performance with SS3T-CSD over MSMT-CSD. Finally, in an age domain-shift setting, SS3T-CSD maintains robust performance across age groups, indicating its potential for more accurate neonatal brain imaging.
Open AccessGround-truth effects in learning-based fiber orientation distribution estimation in neonatal brains
Diffusion Magnetic Resonance Imaging (dMRI) is a non-invasive method for depicting brain microstructure in vivo. Fiber orientation distributions (FODs) are mathematical representations extensively used to map white matter fiber configurations. Recently, FOD estimation with deep neural networks has seen growing success, in particular, those of neonates estimated with fewer diffusion measurements. These methods are mostly trained on target FODs reconstructed with multi-shell multi-tissue constrained spherical deconvolution (MSMT-CSD), which might not be the ideal ground truth for developing brains. Here, we investigate this hypothesis by training a state-of-the-art model based on the U-Net architecture on both MSMT-CSD and single-shell three-tissue constrained spherical deconvolution (SS3T-CSD). Our results suggest that SS3T-CSD might be more suited for neonatal brains, given that the ratio between single and multiple fiber-estimated voxels with SS3T-CSD is more realistic compared to MSMT-CSD. Additionally, increasing the number of input gradient directions significantly improves performance with SS3T-CSD over MSMT-CSD. Finally, in an age domain-shift setting, SS3T-CSD maintains robust performance across age groups, indicating its potential for more accurate neonatal brain imaging.
- Open Access
A-eye: Automated 3D MRI Segmentation and Morphometric Feature Extraction for Eye and Orbit Atlas Construction
This study addresses the need for an automated and accurate 3D segmentation of the healthy human eye and orbit from Magnetic Resonance Images, to allow improved ophthalmic diagnostics and treatments. Past efforts primarily focused on small sample sizes and varied imaging modalities. Here, we leverage a large-scale dataset of T1-weighted MRI of 1245 subjects and the use of the deep learning-based nnU-Net for MR-Eye segmentation tasks. The results showcase robust and accurate segmentations of lens, globe, optic nerve, rectus muscles, and fat. We also present the automated estimation of key ophthalmic biomarkers such as AL and volumetry, while benchmarking correlations between body mass index (BMI) and eye structure volumes. Quality control protocols are introduced through the pipeline to ensure the reliability and clinical relevance of the segmented large-scale data, further enhancing the applicability of our algorithm. Teaser This study advances MR-Eye with automated 3D segmentation of human eye and orbit, biomarker extraction, and a large-scale eye atlas in MRI.
Open AccessA-eye: Automated 3D MRI Segmentation and Morphometric Feature Extraction for Eye and Orbit Atlas Construction
This study addresses the need for an automated and accurate 3D segmentation of the healthy human eye and orbit from Magnetic Resonance Images, to allow improved ophthalmic diagnostics and treatments. Past efforts primarily focused on small sample sizes and varied imaging modalities. Here, we leverage a large-scale dataset of T1-weighted MRI of 1245 subjects and the use of the deep learning-based nnU-Net for MR-Eye segmentation tasks. The results showcase robust and accurate segmentations of lens, globe, optic nerve, rectus muscles, and fat. We also present the automated estimation of key ophthalmic biomarkers such as AL and volumetry, while benchmarking correlations between body mass index (BMI) and eye structure volumes. Quality control protocols are introduced through the pipeline to ensure the reliability and clinical relevance of the segmented large-scale data, further enhancing the applicability of our algorithm. Teaser This study advances MR-Eye with automated 3D segmentation of human eye and orbit, biomarker extraction, and a large-scale eye atlas in MRI.
- Open Access
- 25 April 2024
- Published by Elsevier in Medical Image Analysis
- Vol. 95, 103186
Deep learning microstructure estimation of developing brains from diffusion MRI: A newborn and fetal study
- 25 April 2024
- Published by Elsevier in Medical Image Analysis
- Vol. 95, 103186
Open AccessDeep learning microstructure estimation of developing brains from diffusion MRI: A newborn and fetal study