A5047909906- Image Processing Techniques and Applications
- Digital Holography and Microscopy
- Cell Image Analysis Techniques
- Advanced Fluorescence Microscopy Techniques
- Advanced X-ray Imaging Techniques
- Optical Coherence Tomography Applications
- Optical measurement and interference techniques
- Advanced Optical Imaging Technologies
- Photoacoustic and Ultrasonic Imaging
- Advanced Image Processing Techniques
- AI in cancer detection
- Neural Networks and Applications
- Adversarial Robustness in Machine Learning
- Quantum optics and atomic interactions
- Generative Adversarial Networks and Image Synthesis
- Laser-Matter Interactions and Applications
- Statistical and numerical algorithms
- Molecular Biology Techniques and Applications
- Sparse and Compressive Sensing Techniques
- Advanced Optical Sensing Technologies
- Anatomy and Medical Technology
- Digital Imaging in Medicine
- Optical Imaging and Spectroscopy Techniques
- Near-Field Optical Microscopy
- Cognitive Science and Education Research
University of California, Los Angeles
2021-2024
California NanoSystems Institute
2021-2024
Samueli Institute
2021-2024
State Key Laboratory of Modern Optical Instruments
2019
Zhejiang University
2019
State Key Laboratory on Integrated Optoelectronics
2019
Deep learning-based image reconstruction methods have achieved remarkable success in phase recovery and holographic imaging. However, the generalization of their performance to new types samples never seen by network remains a challenge. Here we introduce deep learning framework, termed Fourier Imager Network (FIN), that can perform end-to-end from raw holograms samples, exhibiting unprecedented external generalization. FIN architecture is based on spatial transform modules process...
Abstract Label‐free super‐resolution (LFSR) imaging relies on light‐scattering processes in nanoscale objects without a need for fluorescent (FL) staining required super‐resolved FL microscopy. The objectives of this Roadmap are to present comprehensive vision the developments, state‐of‐the‐art field, and discuss resolution boundaries hurdles that be overcome break classical diffraction limit label‐free imaging. scope spans from advanced interference detection techniques, where...
Abstract Existing applications of deep learning in computational imaging and microscopy mostly depend on supervised learning, requiring large-scale, diverse labelled training data. The acquisition preparation such image datasets is often laborious costly, leading to limited generalization new sample types. Here we report a self-supervised model, termed GedankenNet, that eliminates the need for or experimental data, demonstrate its effectiveness superior hologram reconstruction tasks. Without...
Autofocusing is a critical step for high-quality microscopic imaging of specimens, especially measurements that extend over time covering large fields view. generally practiced using two main approaches. Hardware-based optical autofocusing methods rely on additional distance sensors are integrated with microscopy system. Algorithmic methods, the other hand, regularly require axial scanning through sample volume, leading to longer times, which might also introduce phototoxicity and...
Digital holography is one of the most widely used label-free microscopy techniques in biomedical imaging. Recovery missing phase information on a hologram an important step holographic image reconstruction. Here we demonstrate convolutional recurrent neural network (RNN) based recovery approach that uses multiple holograms, captured at different sample-to-sensor distances, to rapidly reconstruct and amplitude sample while also performing autofocusing through same network. We demonstrated...
Abstract Traditional histochemical staining of post-mortem samples often confronts inferior quality due to autolysis caused by delayed fixation cadaver tissue, and such chemical procedures covering large tissue areas demand substantial labor, cost time. Here, we demonstrate virtual autopsy using a trained neural network rapidly transform autofluorescence images label-free sections into brightfield equivalent images, matching hematoxylin eosin (H&E) stained versions the same samples. The...
Volumetric imaging of samples using fluorescence microscopy plays an important role in various fields including physical, medical and life sciences. Here we report a deep learning-based volumetric image inference framework that uses 2D images are sparsely captured by standard wide-field microscope at arbitrary axial positions within the sample volume. Through recurrent convolutional neural network, which term as Recurrent-MZ, information from few planes is explicitly incorporated to...
The application of deep learning techniques has greatly enhanced holographic imaging capabilities, leading to improved phase recovery and image reconstruction. Here, we introduce a neural network termed Fourier Imager Network (eFIN) as highly generalizable robust framework for hologram reconstruction with pixel super-resolution autofocusing. Through microscopy experiments involving lung, prostate salivary gland tissue sections Papanicolau (Pap) smears, demonstrate that eFIN superior quality...
Multi-spectral imaging, which simultaneously captures the spatial and spectral information of a scene, is widely used across diverse fields, including remote sensing, biomedical agricultural monitoring. Here, we introduce snapshot multi-spectral imaging approach employing standard monochrome image sensor with no additional filters or customized components. Our system leverages inherent chromatic aberration wavelength-dependent defocusing as natural source physical encoding information; this...
Deep learning-based virtual staining was developed to introduce image contrast label-free tissue sections, digitally matching the histological staining, which is time-consuming, labor-intensive, and destructive tissue. Standard requires high autofocusing precision during whole slide imaging of tissue, consumes a significant portion total time can lead photodamage. Here, we fast framework that stain defocused autofluorescence images unlabeled achieving equivalent performance in-focus images,...
We present a virtual refocusing method over an extended depth of field (DOF) enabled by cascaded neural networks and double-helix point-spread function (DH-PSF). This network model, referred to as W-Net, is composed two generator discriminator pairs. The first learns virtually refocus input image onto user-defined plane, while the second perform cross-modality transformation, improving lateral resolution output image. Using this W-Net model with DH-PSF engineering, we experimentally DOF...
Deep learning-based methods in computational microscopy have been shown to be powerful but general face some challenges due limited generalization new types of samples and requirements for large diverse training data. Here, we demonstrate a few-shot transfer learning method that helps holographic image reconstruction deep neural network rapidly generalize using small datasets. We pre-trained convolutional recurrent on dataset with samples, which serves as the backbone model. By fixing blocks...
In OCT angiography (OCTA), decorrelation computation has been widely used as a local motion index to identify dynamic flow from static tissues, but its dependence on SNR severely degrades the vascular visibility, particularly in low- regions. To mathematically characterize decorrelation-SNR of signals, we developed multi-variate time series (MVTS) model. Based model, derived universal asymptotic linear relation inverse (iSNR), with variance and noise regions determined by average kernel...
Fluorescence lifetime imaging microscopy (FLIM) measures fluorescence lifetimes of fluorescent probes to investigate molecular interactions. However, conventional FLIM systems often requires extensive scanning that is time-consuming. To address this challenge, we developed a novel computational technique called light field tomographic (LIFT-FLIM). Our approach acquires volumetric images in highly data-efficient manner, significantly reducing the number steps. We demonstrated LIFT-FLIM using...
Abstract Fluorescence lifetime imaging microscopy (FLIM) is a powerful technique that enables the visualization of biological samples at molecular level by measuring fluorescence decay rate fluorescent probes. This provides critical information about interactions, environmental changes, and localization within systems. However, creating high-resolution maps using conventional FLIM systems can be challenging, as it often requires extensive scanning significantly lengthen acquisition times....
Fluorescence lifetime imaging microscopy (FLIM) is a powerful technique that enables the visualization of biological samples at molecular level by measuring fluorescence decay rate fluorescent probes. This provides critical information about interactions, environmental changes, and localization within systems. However, creating high-resolution maps using conventional FLIM systems can be challenging, as it often requires extensive scanning significantly lengthen acquisition times. issue...
Holographic imaging plays an essential role in label-free microscopy techniques, and the retrieval of phase information a specimen is vital for image reconstruction holography. Here, we demonstrate recurrent neural network (RNN) based holographic methods that simultaneously perform autofocusing from multiple holograms captured at different sample-to-sensor distances. The acquired input are individually back propagated to common axial plane without any retrieval, then fed into trained RNN...
The recovery of obscured objects is an important goal in imaging that has been approached by exploiting coherence properties, ballistic photons, and penetrating wavelengths. In this paper, a robust reconstruction non-line-of-sight (NLOS) algorithm was proposed based on the Bayesian statistics, using temporal, spatial, intensity information each signal. Compared with conventional back-projection methods, able to handle random errors data image occluded higher quality. An adjustable...
We introduce GedankenNet, a self-supervised learning model for hologram reconstruction. During its training, GedankenNet leveraged physics-consistency loss informed by the physical forward of imaging process, and simulated holograms generated from artificial random images with no correspondence to real-world samples. After this experimental-data-free training based on "Gedanken Experiments", successfully generalized experimental first exposure data, reconstructing complex fields various This...