A5027828897- Physics of Superconductivity and Magnetism
- Radiomics and Machine Learning in Medical Imaging
- AI in cancer detection
- Advanced Image Processing Techniques
- Image Processing Techniques and Applications
- Iron-based superconductors research
- Inorganic Fluorides and Related Compounds
- Superconducting Materials and Applications
- Advanced Neural Network Applications
- COVID-19 diagnosis using AI
- Advanced Condensed Matter Physics
- Medical Image Segmentation Techniques
- Digital Imaging for Blood Diseases
- Lung Cancer Diagnosis and Treatment
- EEG and Brain-Computer Interfaces
- Advanced MRI Techniques and Applications
- Functional Brain Connectivity Studies
- Medical Imaging Techniques and Applications
- Cervical Cancer and HPV Research
- Magnetic and transport properties of perovskites and related materials
- Electronic and Structural Properties of Oxides
- Magnetic properties of thin films
- Medical Imaging and Analysis
- Generative Adversarial Networks and Image Synthesis
- Vaccine Coverage and Hesitancy
Shanghai Jiao Tong University
2021-2025
University of California, Berkeley
2025
Pasadena City College
2025
Shenzhen Terahertz Technology Innovation Research Institute
2024
Beihang University
2019
Beijing Advanced Sciences and Innovation Center
2019
Anhui Medical University
2019
Hefei Institute of Technology Innovation
2019
Linac Coherent Light Source
2014
SLAC National Accelerator Laboratory
2014
Alignment between human knowledge and machine learning models is crucial for achieving efficient interpretable AI systems. However, conventional self-supervised pre-training methods often suffer from low efficiency, as they do not incorporate during the process instead rely mainly on post-hoc alignment techniques. We propose Gaze Pre-Training (GzPT), a novel approach that introduces early with eye gaze information to enhance both efficiency performance of models. By leveraging contrastive...
One-way functions are essential tools for cryptography. However, the existence of one-way is still an open conjecture. By constructing a function with classical bits as input and quantum states output, we prove more rigorously. It provides theoretical guarantees security many cryptographic protocols.
Due to the inherent flexibility of prompting, foundation models have emerged as predominant force in fields natural language processing and computer vision. The recent introduction Segment Anything Model (SAM) signifies a noteworthy expansion prompt-driven paradigm into domain image segmentation, thereby introducing plethora previously unexplored capabilities. However, viability its application medical segmentation remains uncertain, given substantial distinctions between images. In this...
Brain serves as a critical cornerstone of human intelligence, which involves series complex neuropsychological activities that lead to the coordination various functions in brain network. In recent years, network analysis methods based on graph neural networks (GNNs) have attracted increasing attention for identification disorders. However, these generally assume is homogeneous while ignoring its heterogeneity among activities, reflected both connectivity and distinctive functions. To...
Cross-modality synthesis (CMS), super-resolution (SR), and their combination (CMSR) have been extensively studied for magnetic resonance imaging (MRI). Their primary goals are to enhance the quality by synthesizing desired modality reducing slice thickness. Despite promising synthetic results, these techniques often tailored specific tasks, thereby limiting adaptability complex clinical scenarios. Therefore, it is crucial build a unified network that can handle various image tasks with...
Accurate automatic segmentation of medical images typically requires large datasets with high-quality annotations, making it less applicable in clinical settings due to limited training data. One-shot based on learned transformations (OSSLT) has shown promise when labeled data is extremely limited, including unsupervised deformable registration, augmentation and from augmented However, current one-shot methods are challenged by diversity during augmentation, potential label errors caused...