Controlling light and sound to see the world better and make a better world.

So far, using optics and ultrasonics, I contributed to biomedical, metallurgical, and industrial research and development.

Overview

Computational Optics

Single-shot quantitative phase imaging

Polarized LED array microscope for single-shot QPI

Quantitative phase imaging (QPI) provides objective and rich imaging contrast of transparent biological specimens. However, it requires multiple measurements of the same scene, limiting the imaging speed. We proposed a polarized LED array microscope for single-shot QPI using polarization-encoded illumination multiplexing. We showcased our method by capturing slow-motion videos of fast-moving microorganisms at a frame rate limited only by the sensor's readout speed.

Optical diffraction tomography (a.k.a. holotomography)

Cellular organelles in human liver cells (Hep3B) captured by holotomography (HT-X1)

Optical diffraction tomography, also known as holotomography, is a 3D quantitative phase imaging technique that uses refractive index (RI) distribution as intrinsic imaging contrasts for label-free imaging. I have contributed crucially to launching HT-X1, the world's first holotomography system using LED illumination. My main contribution was developing a regularized reconstruction algorithm to solve the "missing cone problem" in 3D RI imaging.

Super-resolution holotomography using deep learning

Super-resolution holotomography of human liver cancer cells (HepG2)

Due to the diffraction-limited frequency coverage of holotomography, RI is underestimated and cells appear elongated in the RI tomogram. This problem can be mitigated using regularized reconstruction algorithms for filling in the missing information. However, conventional algorithms have limitations for low-NA tomograms because the amount of unknown information to be inferred is too large compared to that of the known information. In order to solve this problem, we developed a deep-learning algorithm to transform low-NA tomograms into high-NA tomograms.

Non-destructive 3D quantitative imaging of micro-cracks

A captured 3D RI image of a micro-crack formed on an ultrathin glass by a Vickers indenter.

We have expanded holotomography's application to non-destructively inspect cracks in industrial glass samples. We were able to image RI tomograms of micro-cracks and their propagation inside the glass. Moreover, we were able to capture slight differences in RI near the cracks. We launched an industrial HT system (HT-X1i) with this technology HT-X1i) for defect inspection.

Wavefront shaping

Wavefront shaping precompensates the input wavefront in order to deal with multiple scattering. With the help of my experiences in phase conjugation and time reversal in medical ultrasound, I participated in a review paper that thoroughly covers the principles and recent advances of wavefront shaping in the optical realm.

Single-shot polarization-sensitive Fourier ptychographic microscopy (in progress)

Scene-matched lensless imaging (in progress)

& more to come!

Diagnostic/Therapeutic Ultrasound

Source separation algorithm for ultrasound harmonic imaging

Harmonic separation. (a) Raw data, (b) maximum likelihood, (c) posterior, (d) separated harmonic.

Ultrasound harmonic imaging is widely used to obtain images with high contrast and resolution. However, the prevalent pulse-inversion technique for harmonic imaging requires two Tx/Rx sequences in order to separate the nonlinear harmonic signal. We developed an adaptive harmonic-separation technique, which requires only one Tx/Rx sequence, exploiting expectation-maximization source separation. 

Sparse 2D transducer arrays for 3D ultrasound imaging

(Banner image: beam pattern of a sparse spiral array)

Sparse array pairs. (a,b,c) Sparse rectangular arrays, (d,e,f) sparse spiral arrays.

To achieve artifact-free 3D imaging, densely packed 2D arrays are needed. However, the hardware and computational costs are massive with dense arrays. We designed generalized models for sparse 2D arrays that have minimal grating-lobe artifacts.

Noise elimination algorithm for ultrasound-guided HIFU treatment

Ultrasound imaging is a cost-effective tool for monitoring high-intensity focused ultrasound (HIFU) treatment in real time. However, because of the reflected HIFU signal, images are corrupted with HIFU interference. In order to eliminate the HIFU interference in the monitoring images, we developed an imaging algorithm using a sequence of four different transmission signals to cancel out the HIFU noise and its harmonics.

Ultrasonic brain therapy

Ultrasound allows us to noninvasively deliver high enough energy for a targeted therapy inside the brain. As localized transmission is crucial for brain therapy, we proposed a hemispherical transducer array design with low grating-lobe artifacts by optimizing the array pattern, element size, and curvature. Based on the optimal design, an MR-compatible transducer array was manufactured.

In addition, to compensate for the phase aberration induced by the highly heterogeneous skull, we developed time-reversal simulators with k-wave toolbox and COMSOL Multiphysics.  

Pulsed Laser-Assisted Metal 3D Printing

In-situ porosity reduction and grain refinement during metal 3D printing

Metal 3D printing suffers from anisotropic grain structure and porosity defects. In order to solve these problems, a pulsed-laser technique for reducing porosity defects and refining the grain structure during 3D printing was proposed and verified experimentally and theoretically. By focusing the pulsed laser onto the melt pool, accelerated Marangoni flow, shock waves, and cavitation were generated. With these pulsed laser-induced effects, gas traps were pushed away and a favorable environment for the crystallization of equiaxed grains was formed.

Check out my Inspirations.