A5040430392- Medical Image Segmentation Techniques
- 3D Shape Modeling and Analysis
- Robotics and Sensor-Based Localization
- Functional Brain Connectivity Studies
- Neural dynamics and brain function
- Matrix Theory and Algorithms
- Advanced Numerical Analysis Techniques
- Explainable Artificial Intelligence (XAI)
- Adversarial Robustness in Machine Learning
- Rough Sets and Fuzzy Logic
- Advanced Vision and Imaging
- Language, Metaphor, and Cognition
- Action Observation and Synchronization
- Natural Language Processing Techniques
- Electromagnetic Scattering and Analysis
- Computer Graphics and Visualization Techniques
- Data Mining Algorithms and Applications
- Image Retrieval and Classification Techniques
- Optical measurement and interference techniques
- Neuroscience and Neuropharmacology Research
- Photoreceptor and optogenetics research
- Language and cultural evolution
- Spectral Theory in Mathematical Physics
- Computational Geometry and Mesh Generation
- Advanced Image and Video Retrieval Techniques
National Institute of Mental Health
2022-2025
National Institute on Drug Abuse
2025
National Institutes of Health
2025
Queen Mary Hospital
2024
University of Hong Kong
1997-2024
National Institute of Health
2021
California Institute of Technology
2019
Chinese University of Hong Kong
2011-2017
Surface registration between cortical surfaces is crucial in medical imaging for performing systematic comparisons brains. Landmark-matching that matches anatomical features, called the sulcal landmarks, often required to obtain a meaningful 1-1 correspondence brain surfaces. This commonly done by parameterizing surface onto simple parameter domain, such as unit sphere, which landmarks are consistently aligned. can then be obtained from landmark aligned parameterizations. For genus-0 closed...
Surface parameterizations and registrations are important in computer graphics imaging, where 1-1 correspondences between meshes computed. In practice, surface maps usually represented stored as three-dimensional coordinates each vertex is mapped to, which often requires lots of memory. This causes inconvenience data transmission storage. To tackle this problem, we propose an effective algorithm for compressing homeomorphisms using Fourier approximation the Beltrami representation. The...
We present a new approach to obtain diffeomorphic registrations with large deformations using landmark and intensity information via quasi-conformal maps. The basic idea is minimize an energy functional involving Beltrami coefficient term, which measures the distortion of map. effectively controls bijectivity smoothness registration. In this paper, we first propose registration (QCLR) algorithm (1-1 onto) between images or surfaces. Using QCLR, landmark-aligned diffeomorphisms surfaces can...
Registration, which aims to find an optimal 1-1 correspondence between shapes, is important process in different research areas. Landmark-based surface registration has been widely studied obtain a mapping shapes that matches features. Obtaining unique and bijective features consistently generally challenging, especially when large number of landmark constraints are enforced. This motivates us search for matching diffeomorphism, minimizes the local geometric distortion. For this purpose, we...
In this paper, we propose an adaptive fast solver for a general class of symmetric positive definite (SPD) matrices which include the well-known graph Laplacian. We achieve by developing operator compression scheme and multiresolution matrix factorization algorithm nearly optimal performance on both complexity well-posedness. To develop our methods, first introduce novel notion energy decomposition SPD $A$ using representation elements. The interaction between these elements depicts...
Learning when to initiate or withhold actions is essential for survival and requires integration of past experiences with new information adapt changing environments. While stable prelimbic cortex (PL) ensembles have been identified during reward learning, it remains unclear how they contingencies shift. Does the same ensemble adjust its activity support behavioral suppression upon omission, a distinct recruited this learning? We used single-cell calcium imaging longitudinally track PL...
Abstract Revealing the connectivity of functionally identified individual neurons is necessary to understand how activity patterns emerge and support behaviour. Yet brain-wide presynaptic wiring rules that lay foundation for functional selectivity remain largely unexplored. Cortical neurons, even in primary sensory cortex, are heterogeneous their selectivity, not only stimuli but also multiple aspects Here, investigate underlying pyramidal behavioural state 1–10 somatosensory cortex (S1), we...
Whole-brain functional connectivity ( FC ) measured with MRI (fMRI) evolves over time in meaningful ways at temporal scales going from years (e.g., development) to seconds [e.g., within-scan time-varying tvFC )]. Yet, our ability explore is severely constrained by its large dimensionality (several thousands). To overcome this difficulty, researchers often seek generate low dimensional representations 2D and 3D scatter plots) hoping those will retain important aspects of the data...
Analyzing the deformation pattern of an object is crucial in various fields, such as computer vision and medical imaging. A can be considered a combination local global deformations at different locations. To fully understand analyze pattern, extracting components scales locations necessary. We propose algorithm for multiscale decomposition bijective using quasi-conformal theories. described orientation-preserving homeomorphism two-dimensional domain. The mapping then represented by its...
In this paper we propose a new iterative method to hierarchically compute relatively large number of leftmost eigenpairs sparse symmetric positive matrix under the multiresolution operator compression framework. We exploit well-conditioned property every decomposition component by integrating framework into implicitly restarted Lanczos method. achieve combination proposing an extension-refinement scheme, in which intrinsic idea is decompose target spectrum several segments such that...
<p style='text-indent:20px;'>Image registration has been widely studied over the past several decades, with numerous applications in science, engineering and medicine. Most of conventional mathematical models for large deformation image rely on prescribed landmarks, which usually require tedious manual labeling. In recent years, there a surge interest use machine learning registration. this paper, we develop novel method by fusion quasiconformal theory convolutional neural network...
Registration, which aims to find an optimal 1-1 correspondence between shapes, is important process in different research areas. Conformal mappings have been widely used obtain a diffeomorphism shapes that minimizes angular distortion. registrations are beneficial since it preserves the local geometry well. However, when landmark constraints enforced, conformal generally do not exist. This motivates us look for unique matching quasi-conformal registration, conformality Under suitable...
Deep neural networks (DNNs) are being increasingly used to make predictions from functional magnetic resonance imaging (fMRI) data. However, they widely seen as uninterpretable “black boxes,” it can be difficult discover what input information is by the DNN in process, something important both cognitive neuroscience and clinical applications. A saliency map a common approach for producing interpretable visualizations of relative importance features prediction. methods creating maps often...
Neuronal connections provide the scaffolding for neuronal function. Revealing connectivity of functionally identified individual neurons is necessary to understand how activity patterns emerge and support behavior. Yet, brain-wide presynaptic wiring rules that lay foundation functional selectivity remain largely unexplored. Cortical neurons, even in primary sensory cortex, are heterogeneous their selectivity, not only stimuli but also multiple aspects Here, investigate underlying pyramidal...
This paper presents two algorithms, based on conformal geometry, for the multi-scale representations of geometric shapes and surface morphing. A representation aims to describe a 3D shape at different levels detail, which allows analyzing or editing surfaces global local scales effectively. Surface morphing refers process interpolating between shapes, has been widely applied estimate analyze deformations in computer graphics, vision medical imaging. In this work, we propose models surfaces....
Surface parameterizations and registrations are important in computer graphics imaging, where 1-1 correspondences between meshes computed. In practice, surface maps usually represented stored as 3D coordinates each vertex is mapped to, which often requires lots of storage memory. This causes inconvenience data transmission storage. To tackle this problem, we propose an effective algorithm for compressing homeomorphisms using Fourier approximation the Beltrami representation. The...
In order to spread malware more effectively, hackers have started target popular social networking services (SNS) due the inherent trust-relationship between SNS users and interactive nature of services. A common attacking approach is for a automatically login using stolen user credentials then deliver malicious weblinks (Uniform Resource Locators (URLs)) people on contact/friend-list account by embedding them in some short messages. The victim gets infected clicking links thought be...
Abstract Whole-brain functional connectivity ( FC ) measured with MRI (fMRI) evolve over time in meaningful ways at temporal scales going from years (e.g., development) to seconds within-scan time-varying tvFC )). Yet, our ability explore is severely constrained by its large dimensionality (several thousands). To overcome this difficulty, researchers seek generate low dimensional representations 2D and 3D scatter plots) expected retain most informative aspects relationships behavior, disease...