PROJECTS

Biomarker Extraction for Ocular Diseases

2018 - present

Collaborated with multiple research institutions to collect clinical data for identifying biomarkers associated with AMD. Imaging features include drusen, geographic atrophy, calcified drusen, choriocapillaris flow deficits, outer retinal layer thickness, hyperreflective foci, choroidal thickness etc. Developed machine learning based segmentation algorithms for each biomarker and analyzed the risk factor for the development of late AMD for NIH natural history study grants.

Extending the Capabilities of Current OCT Systems

2017 - 2021

Improved contrast, resolution, and extended the field of view (FOV) for current commercial OCT systems. Analyzed the distortion model in the eye and implemented automated stitching to extend the FOV to 80 degrees. Developed a 3D image registration to improve signal-to-noise ratio and visualize subtle structures under 10um, such as photoreceptors and choriocapillaris.

Assessment and Quantification of Skin Properties

2018-2020

Used non-invasive 3D imaging techniques to quantify human skin, evaluating properties such as thickness, roughness, and capillary morphology. Applied the evaluation pipeline to patients with acute burn injuries. Managed collaborations with Shiseido and Estee Lauder for assessing aging and cosmetic products.

Deep Learning for clinical image diagnosis 

2017 - 2018

Design and apply deep convolution network for identifying and evaluating retinal pathologies with limited clinical retinal optical coherence tomography (OCT) images. 

Fast 3-D registration for OCT speckle reduction and visualization

2018 - 2020 

Propose an approach for fast 3-D registration and speckle reduction for OCT and OCT-Angiography which can dramatically increase the contrast of images and suppress the motion induced artifacts. 

Statistical analysis in optical coherence tomography angiography

December 2014 - June 2016

Developed a mathematical description of the statistical properties of the OCT angiography. This statistical model is helpful to furtherly understand the origin of the motion contrast in OCT angiography, quantitatively predicate the imaging performance, and guide the optimization of the system and the associated algorithm.

Development of a multifunctional microscope for multiscale imaging

April 2015 - June 2016

The goal of this project is to develop a new microscope that performs multiscale study of cortical function using intrinsic signal imaging, OCT, and multiphoton imaging in vivo. This will permit study of functional organization, neuronal response, laminar visualization, and vascular mapping of the same cortical area. 

Quantification of label-free OCT angiography

October 2015 - June 2016

Thesis project of bachelor degree. I applied the label-free OCT-A technique to quantify the vascular occlusion and evaluate the changes in vessel structural information in rats cortex during the stroke progression.

GANs for image completion

May 2017

Explore the application of the generative adversarial networks (GANs) model in image completion and also in the recovery of noisy medical images which hasn't been explored to our best knowledge. 

Face completion: to recover partially covered face, for applications like picture beautification or criminal identification in videos. 

Medical image noise completion: to improve image quality by by removing noises, e.g., retinal angiograms with noise brought in during image acquisition. 

Data mining for public transport in Seattle

November 2016

Collected public bicycles (Pronto) and  carsharing (Car2go) data to analysis the pattern of public transport in Seattle and provide a better travel strategy for the companies and users.

Automatic gobang system

January 2016

Designed and fabricated an automatic gobang system, which can recognize the chess pieces by computer vision algorithm and automatic place the pieces by its AI and electromechanical system.