Embedded Files
I recognize my academic contributions in two fields, mathematics and automation. My master thesis (later a journal) is about discovering of an analytic solution to exponentially distributed trap carrier transport model, in which space-charge-limited (SCL), alternative to drift, currents are dominating. Because the expression of the current is analytic, it is found that the current is more sensitive to temperature than that predicted by its asymptotic expression.The model may be suitable for applications in Organic LEDs. In PhD, I developed a novel optical imaging system termed angled fluorescence laminar optical tomography (aFLOT). aFLOT was applied in the field of tissue engineering. As a non-invasive imaging tool, it resolves 3D cell (hMSC) distributions in large constructs at milli-meter scale. It fills the gap where high resolution microscopy cannot image deep and macroscopy (such as CT, MRI) cannot resolve high enough to cellular level.
Education
Education
Straight Electrical Engineering:
- BS in 2005 at National Taiwan University.
- MS in 2007 at National Taiwan University, semiconductor
- PhD in 2013 at the University of Maryland, College park, electrophysics, medical imaging.
Work experience
Work experience
In summer 2004, I interned at Genesys Logic, Inc. where I designed a circuit board to evaluate its high speed commercial analog-to-digital converter. To finance my PhD education, I interned at Signal Processing, Inc. (SPI) in Spring 2008 where I implemented independent component analysis (ICA) to handle the so-called blind source separation (BSS) problems to improve astronauts' communication, a project funded by NASA. My PhD adviser Dr. Yu Chen secured me a contractor position at Food and Drug Administration (FDA) from Fall 2011 to Fall 2012. Co-advised by Dr. Joshua Pfefer, lab leader, CDRH, FDA, I developed phantoms, metrics, and methods to evaluate the performance among various OCT machines. Upon graduation in May 2013, I moved to Bay area. I am currently a system software engineer at KLA-Tencor.
Computer skills
Computer skills
- 2002-2013, Matlab was my major programming language. once actively in the community. Matlab-rize everything.
- Since 2013, Java is now my major language for work (machine control). I also program Java for automation using Sikuli, image processing (ImageJ), and web applications (DropWizard)
- Since 2013, Python is my awesome Matlab alternative. See me at meetups!
Interests
Interests
Automation
Automation
Several instances during my education motivated me to pursue machine automation. I routinely needed to measure the current of OLEDs. While the cheap solution can always be to operate instruments manually, it is never one when fast, batch, systematic measurement becomes necessary. To accelerate the process, I interfaced Keithley 2400 with Matlab and successfully automated the measurements. At UMCP, I applied the same trick again to automate a zapping process, a process that gradually increases current to "burn" the conducting (current leaking) part of carbon nanotube until burned. When building the aFLOT system, I also needed to coordinate the translation stage and camera. I developed the stage-controlling macro, interfaced EMCCD, and used PC as the centralized coordinator among subsystems.
I consciously realize that labor intensive routine processes are commonly present in academia, industries, and daily lives, even today. Formulating efficient process to handle routines is a pattern of my action. It saves time in the long run. Identifying routine also makes me a keen observer.
Mathematics
Mathematics
NEW BOUNDS FOR SPHERICAL TWO-DISTANCE SETS
Image processing
Image processing
3D labeling
3D labeling is a process that works on the binarized volume and labels isolated regions. I once developed my own version of 3d labeling process based on 2d labeling, (which is based on 1d labeling, if you walk through the logic,) but later I realized this is an established union-find problem. Matlab had built in bwlabeln, so had other tomography processing tools. Nevertheless, it was a rewarding learning experience. The figure on the right is a sample done by my naive program.
3D partitioning
3D partitioning is a process that works on the binarized volume and partitions weakly connected regions, a step further than the 3D labeling. I developed one routine, which does 3D skeletonization, analyzes the skeleton, classifies skeletons (to determine if breakable), labels the skeletons, and dilates the skeletons back to volumes. Figure on the right shows an example of an originally 3-connected pores being partitioned as well as the "skeleton".
Structured illumination
Optics
Optics
Angled Fluorescence Laminar Optical Tomography (aFLOT)
My PhD thesis is to develop a new mesoscopic tomography, named angled fluorescence laminar optical tomography (aFLOT). Figure demonstrates human mesenchymal stem cell (hMSC) distribution resolved by aFLOT, FOV: 6.5 x 10.6 x 2.9 mm3.
aFLOT is within a family of fluorescence molecular tomography, focusing on several mm in depth field of view and tens of micron resolution. It captures fluorescence emitted from the surface of (possibly heterogeneous) sample and models photon distribution within the sample to estimate where the fluorescing sources are. The purpose of the modeling is in an attempt to perform PSF deconvolution to better resolve features. Conventional FLOT and also other FMTs have practiced the same idea. aFLOT further improves resolution and optical sectioning by introducing angle incidence. For more info, please see here.
Characterization of Optical Coherence Tomography (OCT)
This work presents one of FDA's interests in regulating medical devices. To characterize the OCT resolution, our approach is to embed sub-resolution particles into phantoms. As such, the captured image approximates point spread function (PSF) of the corresponding OCT system. The approach has also been practiced in confocal microscopy and two-photon microscopy before being applied in OCT.The advantage is the capability to perform point spread engineering, which is to use the captured PSF as feedback to modify system characteristics to optimize the shape of the PSF, thus improving system resolution. The purpose of the current study was to advance understanding of the capabilities and best practices for use of PSF phantoms and to provide novel, quantitative insights into the performance of OCT system performance.
Top figure shows one OCT performance using our metrics: FWHM of PSF in x y z directions and the peak intensity of PSFs along depths. 3 sample PSFs at different depths are shown for visual comparison. Bottom figure compares 3 different OCT performance using the same metrics.
In this project, I developed the phantom and the methods to derive FWHMs and intensities from the acquired OCT image volume.
Optical Coherence Tomography/Elastography/Endoscopy
Modifying OCT to suit medical applications was my side projects. I routinely assisted our lab's leading research in immediate evaluation of kidney function after kidney transplantation surgery using our hand-held OCT device.
Google Sites
Report abuse