The Zhao Lab
Stephenson School of Biomedical Engineering
University of Oklahoma, Norman campus
Computationally-enhanced molecular spectroscopic imaging
Visualizing and quantifying molecular information within subcellular structures in their native states provides new insights into complex biological processes but remains a grand challenge. To address this challenge, we are developing bio-imaging platforms that integrate optical mid-infrared photothermal imaging, quantitative phase imaging, and fluorescence imaging. Our goal is to achieve volumetric, quantitative chemical imaging with high spatial and temporal resolution, as well as 3D IR spectroscopic analysis capabilities.​​
Probe beam scanning
Mid-IR pump beam
Camera
Tube lens
Probe light
Bond-Selective Intensity Diffraction Tomography (BS-IDT) system
Label-free pump-probe imaging based on mid-IR photothermal effects.
3D label-free chemical imaging of cancer cells and C. elegans. Blue: proteins Red: lipids
​​Recently, we have demonstrated the first non-interferometric mid-IR photothermal diffraction tomography imaging and its applications in 3D mapping bladder cancer cell metabolic profiles and C. elegans (Nature Communications 13: 7767, 2022). Our technique also demonstrates the same spectral fidelity as Fourier Transform Infrared Spectroscopy (FTIR) and is superior to FTIR due to its depth-resolved spectroscopy capabilities.​
We further showcase the integration of this technique with fluorescence guidance to visualize tau fibrils’ beta-sheet 3D spatial distributions within intact cells, highlighting its potential application in studying neurodegenerative diseases (Light: Science & Applications 12: 147, 2023).
3D visualization of β sheet distributions within intracellular tau fibrils
BS-IDT mid-IR fingerprint spectra
Images are adapted from our publication, Nature Communications, 13, 7767 (2022).
Back to Top
Artificial-intelligence-enhanced fiber-optic imaging
Fiber-optic endoscopy enables imaging deep within hollow organs or tissues, serving an essential function in both clinical practice and fundamental research. Deep-learning-based fiber-optic imaging solutions have gained popularity due to their outstanding capability to recover high-fidelity images from fiber-delivered degraded images or even scrambled speckle patterns with minimized penetration damage. To achieve superior fiber-optic imaging performance, we are working on developing advanced algorithms and novel optical fiber devices.​​
Cross-section of Glass-Air Anderson Localizing Optical Fiber (GALOF)
Monochromatic fiber-optic cell imaging based on a supervised learning method
Images are adapted from our publication, Advanced Photonics, 1 (6), 066001 (2019).
Previously, we developed Glass-Air Anderson Localizing Optical Fiber (GALOF) to enhance imaging quality and robustness, along with a supervised learning-based GALOF imaging system to enable robust and nearly artifact-free image transport (Optica 5: 984-987, 2018, Advanced Photonics 1: 066001, 2019). Recently, we further developed an unsupervised learning-based GALOF imaging system by integrating a customized cycle generative adversarial network with our GALOF (Light: Science & Applications 12: 125, 2023). This new method demonstrated that unpaired small training data can achieve robust, artifact-free transport of full-color biological images over a meter-long distance under both transmission and reflection modes, as well as remarkable cross-domain generalizability.​
Back to Top
Full-color, fiber-optic, robust cell imaging based on a semi-supervised learning method
Images are adapted from our publication, Light: Science & Applications 12: 125 (2023).
