A5058564351- Cell Image Analysis Techniques
- Image Processing Techniques and Applications
- Neurobiology and Insect Physiology Research
- Advanced Electron Microscopy Techniques and Applications
- Neural dynamics and brain function
- Advanced Fluorescence Microscopy Techniques
- Vestibular and auditory disorders
- Hearing, Cochlea, Tinnitus, Genetics
- Fossil Insects in Amber
- Olfactory and Sensory Function Studies
- AI in cancer detection
- Electron and X-Ray Spectroscopy Techniques
- Circadian rhythm and melatonin
- Digital Holography and Microscopy
- Plant and animal studies
- Machine Learning in Materials Science
- Neuroscience and Neuropharmacology Research
- Domain Adaptation and Few-Shot Learning
- Digital Imaging for Blood Diseases
- Medical Image Segmentation Techniques
- Machine Learning in Bioinformatics
- Mitochondrial Function and Pathology
- Pancreatic function and diabetes
- Machine Learning and Data Classification
- Memory and Neural Mechanisms
Salk Institute for Biological Studies
2021-2024
Janelia Research Campus
2019-2023
Howard Hughes Medical Institute
2018-2020
We present a method combining affinity prediction with region agglomeration, which improves significantly upon the state of art neuron segmentation from electron microscopy (EM) in accuracy and scalability. Our consists 3D U-NET, trained to predict affinities between voxels, followed by iterative agglomeration. train using structured loss based on MALIS, encouraging topologically correct segmentations obtained thresholding. extension two parts: First, we quasi-linear compute gradient,...
Cells respond to mitochondrial poisons with rapid activation of the adenosine monophosphate-activated protein kinase (AMPK), causing acute metabolic changes through phosphorylation and prolonged adaptation metabolism transcriptional effects. Transcription factor EB (TFEB) is a major effector AMPK that increases expression lysosome genes in response energetic stress, but how activates TFEB remains unresolved. We demonstrate directly phosphorylates five conserved serine residues...
Abstract The development of high-resolution microscopes has made it possible to investigate cellular processes in 3D and over time. However, observing fast dynamics remains challenging because photobleaching phototoxicity. Here we report the implementation two content-aware frame interpolation (CAFI) deep learning networks, Zooming SlowMo Depth-Aware Video Frame Interpolation, that are highly suited for accurately predicting images between image pairs, therefore improving temporal resolution...
Abstract We present an auxiliary learning task for the problem of neuron segmentation in electron microscopy volumes. The consists prediction local shape descriptors (LSDs), which we combine with conventional voxel-wise direct neighbor affinities boundary detection. capture statistics about to be segmented, such as diameter, elongation, and direction. On a study comparing several existing methods across various specimen, imaging techniques, resolutions, LSDs consistently increases accuracy...
Abstract The study of neural circuits requires the reconstruction neurons and identification synaptic connections between them. To scale to size whole-brain datasets, semi-automatic methods are needed solve those tasks. Here, we present an automatic method for partner in insect brains, which uses convolutional networks identify post-synaptic sites their pre-synaptic partners. can be trained from human generated point annotations alone require only simple post-processing obtain final...
ABSTRACT Neural representations of head direction have been discovered in many species. A large body theoretical work has proposed that the dynamics associated with these is generated, maintained, and updated by recurrent network structures called ring attractors. We performed electron microscopy-based circuit reconstruction RNA profiling identified cell types heading system Drosophila melanogaster to directly determine underlying neural network. motifs hypothesized maintain representation...
Abstract We present a simple, yet effective, auxiliary learning task for the problem of neuron segmentation in electron microscopy volumes. The consists prediction Local Shape Descriptors ( LSDs ), which we combine with conventional voxel-wise direct neighbor affinities boundary detection. shape descriptors are designed to capture local statistics about be segmented, such as diameter, elongation, and direction. On large study comparing several existing methods across various specimen,...
Through digital imaging, microscopy has evolved from primarily being a means for visual observation of life at the micro- and nano-scale, to quantitative tool with ever-increasing resolution throughput. Artificial intelligence, deep neural networks, machine learning are all niche terms describing computational methods that have gained pivotal role in microscopy-based research over past decade. This Roadmap is written collectively by prominent researchers encompasses selected aspects how...
ABSTRACT Producing dense 3D reconstructions from biological imaging data is a challenging instance segmentation task that requires significant ground-truth training for effective and accurate deep learning-based models. Generating intense human effort to annotate each of an object across serial section images. Our focus on the especially complicated brain neuropil, comprising extensive interdigitation dendritic, axonal, glial processes visualized through electron microscopy. We developed...
The development of high-resolution microscopes has made it possible to investigate cellular processes in 4D (3D over time). However, observing fast dynamics remains challenging as a consequence photobleaching and phototoxicity. These issues become increasingly problematic with the depth volume acquired speed biological events interest. Here, we report implementation two content-aware frame interpolation (CAFI) deep learning networks, Zooming SlowMo (ZS) Depth-Aware Video Frame Interpolation...
Summary The cerebellum is thought to detect and correct errors between intended executed commands 1–3 critical for social behaviors, cognition emotion 4–6 . Computations motor control must be performed quickly in real time should sensitive small differences patterns fine error correction while being resilient noise 7 Influential theories of cerebellar information processing have largely assumed random network connectivity, which increases the encoding capacity network’s first layer 8–13...
Producing dense 3D reconstructions from biological imaging data is a challenging instance segmentation task that requires significant ground-truth training for effective and accurate deep learning-based models. Generating intense human effort to annotate each of an object across serial section images. Our focus on the especially complicated brain neuropil, comprising extensive interdigitation dendritic, axonal, glial processes visualized through electron microscopy. We developed novel method...
As the field of connectomics strives to tackle questions regarding increasingly large neuronal circuits, technologies improving imaging throughput will be vital. X-Ray Holographic Nanotomography (XNH) may play a key role, by allowing for fast, non-destructive, multi-resolution imaging. XNH is well suited rapidly tissue volumes, with easily increased at cost resolution and image quality. We therefore set out systematically examine potential cycle-consistent generative adversarial networks...
Abstract Large-volume ultrastructural mapping approaches yield detailed circuit wiring diagrams but lack an integrated synaptic activity readout which is essential for functional interpretation of the connectome. Here we resolve this limitation by combining labelling in vivo with focused ion-beam scanning electron microscopy (FIBSEM) and machine learning-based segmentation. Our approach generates high-resolution near-isotropic three-dimensional readouts activated vesicle pools across large...