Phase Recovery And Holographic Image Reconstruction Using Neural Networks

Fast computing timeSmall imaging system form factorLow costWidely applicableRequires less imaging stepsEasy to operate

Any phase recovery problem Electron holographyX-ray imagingDiffraction TomographyQuantitative phase microscopyAll forms of medical imaging

A novel microscopy method was developed to extract quantitative phase images and intensity images from a single hologram. The approach combines a lens-free microscope with a CMOS sensor and a deep neural network training algorithm. Many hologram images, intensity images as well as reconstructed phase images of pathology slides were fed into the deep neural training algorithm to compute the fixed deep neural network of convolutional layers, unsampling blocks and nonlinear activation functions. The fixed deep neural training algorithm outputs intensity and phase images of the pathology slides with extremely fast computing time and incredible accuracy.

Background

Quantitative phase imaging is widely applicable in medical imaging. Conventional microscopy methods have bulky optic setups and require a lot of resources and time to produce images. Newer lens-free microscopy methods cut down the form factor of imaging systems and speed up the imaging process for obtaining regular intensity microscopy images. However, these newer methods involve using sensors such as charge coupled devices (CCDs), or complementary metal-oxide-semiconductor (CMOS). These sensors, though work well for obtaining intensity images, cannot directly record quantitative phase information or light diffraction. Previous works have applied digital holography to resolve these issues, but the approaches require reconstruction of multiple holograms and long computing time.

Additional Technologies by these Inventors

• Ultra-Large Field-of-View Fluorescent Imaging Using a Flatbed Scanner
• Detection and Spatial Mapping of Mercury Contamination in Water Samples Using a Smart-Phone
• Automated Semen Analysis Using Holographic Imaging
• Tunable Vapor-Condensed Nano-Lenses
• Single Molecule Imaging and Sizing of DNA on a Cell Phone
• Rapid, Portable And Cost-Effective Yeast Cell Viability And Concentration Analysis Using Lensfree On-Chip Microscopy And Machine Learning
• Wide-Field Imaging Of Birefringent Crystals In Synovial Fluid Using Lens-Free Polarized Microscopy For Crystal Arthropathy Diagnosis
• Quantitative Fluorescence Sensing Through Highly Autofluorescent And Scattering Media Using Cost-Effective Mobile Microscopy
• Demosaiced Pixel Super-Resolution For Multiplexed Holographic Color Imaging
• Microscopic Color Imaging And Calibration
• Pixel Super-Resolution Using Wavelength Scanning
• High-Throughput And Label-Free Single Nanoparticle Sizing Based On Time-Resolved On-Chip Microscopy
• Fluorescent Imaging Of Single Nano-Particles And Viruses On A Smart-Phone
• Holographic Opto-Fluidic Microscopy
• Lensfree Wide-Field Fluorescent Imaging On A Chip Using Compressive Decoding
• Revolutionizing Micro-Array Technologies: A Microscopy Method and System Incorporating Nanofeatures
• Sparsity-Based Multi-Height Phase Recovery In Holographic Microscopy
• Computational Out-Of-Focus Imaging Increases The Space-Bandwidth Product In Lens-Based Coherent Microscopy
• Computational Sensing Using Low-Cost and Mobile Plasmonic Readers Designed by Machine Learning
• Deep Learning Microscopy
• Mobile Phone Based Fluorescence Multi-Well Plate Reader
• Lensfree Tomographic Imaging
• Air Quality Monitoring Using Mobile Microscopy And Machine Learning

Tech ID/UC Case

29015/2017-782-0

Related Cases

2017-782-0

美國

桌面版