A5025614674- Advanced MRI Techniques and Applications
- Primate Behavior and Ecology
- Functional Brain Connectivity Studies
- Metabolomics and Mass Spectrometry Studies
- Medical Imaging Techniques and Applications
- Generative Adversarial Networks and Image Synthesis
- Anomaly Detection Techniques and Applications
- Digital Media Forensic Detection
- Zebrafish Biomedical Research Applications
- Single-cell and spatial transcriptomics
- Neuroinflammation and Neurodegeneration Mechanisms
- Neural dynamics and brain function
- Advanced Image and Video Retrieval Techniques
- Cell Image Analysis Techniques
- Neural Networks and Applications
- Image Retrieval and Classification Techniques
- Ultrasound Imaging and Elastography
- Machine Learning and ELM
- Cardiovascular Function and Risk Factors
- Data Stream Mining Techniques
- Image Processing Techniques and Applications
- Human Pose and Action Recognition
- Cardiac Fibrosis and Remodeling
- Neurobiology and Insect Physiology Research
- Handwritten Text Recognition Techniques
California Institute of Technology
2022-2025
SIB Swiss Institute of Bioinformatics
2020-2022
ETH Zurich
2020-2022
University of Zurich
2019-2022
Technische Universität Braunschweig
2016
Abstract Non-invasive, molecularly-specific, focal modulation of brain circuits with low off-target effects can lead to breakthroughs in treatments disorders. We systemically inject engineered ultrasound-controllable drug carriers and subsequently apply a novel two-component Aggregation Uncaging Focused Ultrasound Sequence (AU-FUS) at the desired targets inside brain. The first sequence aggregates millimeter-precision by orders magnitude. second uncages carrier’s cargo locally achieve high...
Abstract Cells are a fundamental unit of biological organization, and identifying them in imaging data – cell segmentation is critical task for various cellular experiments. While deep learning methods have led to substantial progress on this problem, most models use specialist that work well specific domains. Methods learned the general notion “what cell” can identify across different domains proven elusive. In work, we present CellSAM, foundation model generalizes diverse data. CellSAM...
The reliability of scientific results critically depends on reproducible and transparent data processing. Cross-subject cross-study comparability imaging in general, magnetic resonance (MRI) particular, is contingent the quality registration to a standard reference space. In small animal MRI this not adequately provided by currently used processing workflows, which utilize high-level scripts optimized for human data, adapt fit scripts, rather than vice-versa. fully article we showcase...
Abstract The quantification of behaviors interest from video data is commonly used to study brain function, the effects pharmacological interventions, and genetic alterations. Existing approaches lack capability analyze behavior groups animals in complex environments. We present a novel deep learning architecture for classifying individual social animal behavior, even environments directly raw frames, while requiring no intervention after initial human supervision. Our behavioral classifier...
Abstract Blood Oxygen Level-Dependent (BOLD) functional Magnetic Resonance Imaging (fMRI) allows for non-invasive, indirect recordings of neural activity across the whole brain in both humans and animals. However, relationship between local population vascular is not completely understood. To investigate this relationship, we present a novel MRI compatible single-photon microscope capable measuring cellular resolution Ca 2+ genetically defined neurons during whole-brain BOLD fMRI awake...
Large-scale research integration is contingent on seamless access to data in standardized formats. Standards enable researchers understand external experiment structures, pool results, and apply homogeneous preprocessing analysis workflows. Particularly, they facilitate these features without the need for numerous potentially confounding compatibility add-ons. In small animal magnetic resonance imaging, an overwhelming proportion of acquired via ParaVision software Bruker Corporation. The...
Cells are the fundamental unit of biological organization, and identifying them in imaging data - cell segmentation is a critical task for various cellular experiments. While deep learning methods have led to substantial progress on this problem, models that seen wide use specialist work well specific domains. Methods learned general notion "what cell" can identify across different domains proven elusive. In work, we present CellSAM, foundation model generalizes diverse data. CellSAM builds...
Summary Artificial activation of anatomically localized, genetically defined hypothalamic neuron populations is known to trigger distinct innate behaviors, suggesting a nucleus-centered organization behavior control. To assess whether the encoding similarly confined, we performed simultaneous recordings across twenty regions in freely moving animals. Here show that but distributed ensembles encode social and fear classes, primarily through mixed selectivity. While class-encoding were...
Abstract The reliability of scientific results critically depends on reproducible and transparent data processing. Cross-subject cross-study comparability imaging in general, magnetic resonance (MRI) particular, is contingent the quality registration to a standard reference space. In small animal MRI this not adequately provided by currently used processing workflows, which utilize high-level scripts optimized for human data, adapt fit scripts, rather than vice-versa. fully article we...
The asymmetric resonance frequency analysis of silicon cantilevers for a low-cost wearable airborne nanoparticle detector (Cantor) is described in this paper. cantilevers, which are operated the fundamental in-plane mode, used as mass-sensitive microbalance. They manufactured out bulk silicon, containing full piezoresistive Wheatstone bridge and an integrated thermal heater reading measurement output signal stimulating excitation, respectively. To optimize sensor performance, with different...
The study of social interactions and collective behaviors through multi-agent video analysis is crucial in biology. While self-supervised keypoint discovery has emerged as a promising solution to reduce the need for manual annotations, existing methods often struggle with videos containing multiple interacting agents, especially those same species color. To address this, we introduce B-KinD-multi, novel approach that leverages pre-trained segmentation models guide scenarios. This eliminates...
Mid-level vision capabilities - such as generic object localization and 3D geometric understanding are not only fundamental to human but also crucial for many real-world applications of computer vision. These abilities emerge with minimal supervision during the early stages visual development. Despite their significance, current self-supervised learning (SSL) approaches primarily designed evaluated high-level recognition tasks, leaving mid-level largely unexamined. In this study, we...
We introduce Discovering Conceptual Network Explanations (DCNE), a new approach for generating human-comprehensible visual explanations to enhance the interpretability of deep neural image classifiers. Our method automatically finds that are critical discriminating between classes. This is achieved by simultaneously optimizing three criteria: should be few, diverse, and human-interpretable. builds on recently introduced Concept Relevance Propagation (CRP) explainability method. While CRP...
Self-supervised learning (SSL) is a machine approach where the data itself provides supervision, eliminating need for external labels. The model forced to learn about structure or context by solving pretext task. With SSL, models can from abundant and cheap unlabeled data, significantly reducing cost of training labels are expensive inaccessible. In Computer Vision, SSL widely used as pre-training followed downstream task, such supervised transfer, few-shot on smaller labeled sets, and/or...
Diffusion models are generative with impressive text-to-image synthesis capabilities and have spurred a new wave of creative methods for classical machine learning tasks. However, the best way to harness perceptual knowledge these visual tasks is still an open question. Specifically, it unclear how use prompting interface when applying diffusion backbones vision We find that automatically generated captions can improve text-image alignment significantly enhance model's cross-attention maps,...
Disentanglement is at the forefront of unsupervised learning, as disentangled representations data improve generalization, interpretability, and performance in downstream tasks. Current approaches remain inapplicable for real-world datasets since they are highly variable their fail to reach levels disentanglement (semi-)supervised approaches. We introduce population-based training (PBT) improving consistency variational autoencoders (VAEs) demonstrate validity this approach a supervised...
Disentanglement is at the forefront of unsupervised learning, as disentangled representations data improve generalization, interpretability, and performance in downstream tasks. Current approaches remain inapplicable for real-world datasets since they are highly variable their fail to reach levels disentanglement (semi-)supervised approaches. We introduce population-based training (PBT) improving consistency variational autoencoders (VAEs) demonstrate validity this approach a supervised...
As the nexus of visual identity for companies, a trademark serves as main platform company’s goods and services. It is important that trademarks be distinguished from one another in and/or (visual) similarity to avoid consumer confusion. We conducted study using two approaches achieve deeper understanding process behind judgments. First, we used deep neural networks (DNNs) assess degree image applications. database provided by Japan Patent Office online competition “AI x Trademark: Image...
Abstract Large-scale research integration is contingent on seamless access to data in standardized formats. Standards enable researchers understand external experiment structures, pool results, and apply homogeneous preprocessing analysis workflows. Particularly, they facilitate these features without the need for numerous potentially confounding compatibility add-ons. In small animal magnetic resonance imaging, an overwhelming proportion of acquired via ParaVision software Bruker...