Albert Liang received a B.S. in Electrical and Computer Engineering from Cornell University in 2007, a M.Eng. in Biomedical Engineering also from Cornell University in 2008, and a Ph.D. in Biomedical Engineering from the University of Michigan in 2018. He worked briefly at M. D. Anderson Cancer Center in Houston, TX as a clinical researcher from 2008-2009. Currently, he is a postdoctoral fellow in Dr. Larry Antonuk's flat-panel imager group at UM, where he performs large-scale circuit simulations to design next-generation x-ray sensors for medical applications. His past research experience also includes numerous years of wet lab experiments in the field of biochemistry performing PCRs, Southern blots, Western blots, and protein purification.
He is a member of the American Association of Physicists in Medicine (AAPM) and a member of the Quill and Dagger Senior Honor Society of Cornell University. In the past, he was also co-president of the UM Postdoctoral Association and a member of the Senate Advisory Committee for University Affairs on the Research Policies Committee. He is a UROP instructor and leads the MATLAB workshop for the program and in the past has been a PIBS 503 (Research Ethics) facilitator. In his free time, he volunteers as a board member for the Intercollegiate Taiwanese American Students Association as a board of directors member and at the Humane Society of Huron Valley, and is also an avid volleyball player.
Photon counting X-ray imagers based on poly-Si
Flat-panel x-ray imagers are commonly used in modern medical imaging. Compared to traditional film, these new imagers allow "live" viewing of the image, continuous imaging (fluoroscopy), and advanced noise-reduction techniques. To produce an image, these imagers integrate the energy of x-ray photons that reach the imager. However, during this integration, noise is also commingled, which affects the quality of image. The primary noise source is due to the readout electronics used to operate these imagers, and the magnitude of this noise scales with the size of the imager. Medical x-ray imagers need to be very large in order to image human anatomy and, therefore, this electronic readout noise can be quite large (on the order of 1000 to 3000 e-). Prospects for substantially reducing this noise are poor and, in order to maintain acceptable signal-to-noise ratio in order to make medical diagnoses, a certain amount of x-ray radiation dose is required to produce an image.
Given that excessive x-ray radiation is undesirable (can increase the risk of cancer), my current research is focused on methods to produce better or similar quality images using lower radiation dose, primarily by improving hardware to increase the amount of signal produced by each x-ray. This is achieved using two approaches: by investigating new types of x-ray converter materials (that are responsible for converting x-ray photons into electrical signal) and by improving the circuits of the x-ray imager (that capture the eletrical signal for readout to a computer). The converter material under investigation is mercuric iodide, and the improved circuit is a photon counting imager that digitized the x-ray signal before readout (and is therefore immune to the detrimental electronic additive noise).
Characterization of poly-Si photon counting imager prototypes
A total of 11 different prototype photon counting arrays (PCAs) have been designed, and several copies of each design have been manufactured. The prototype arrays are currently undergoing testing and measurement. Each array consists of 4 components: an amplifier, a comparator, a clock generator, and a counter. Numerous candidate designs were created for each component, and different combinations of each design were picked in order to make a new array design.
Currently, the various designs for each component are being characterized individually. This is possible due to numerous test circuits that were created alongside the PCAs that expose various internal nodes for probing and/or injecting signals. By individually characterizing each design of a component, we can benchmark the performance of that design without the confounding variables associated with being connected in series with the other components.
Developing new poly-Si circuits using machine learning techniques
The amplifier component is the most important component of the photon counting arrays. It needs to be fairly high speed (on the order of microseconds) and provide high gain (on the order of 100x to 1000x), while introducing as little additive noise as possible. The current "best" amplifier circuit identified from our previous studies is a 3-stage folded cascode amplifier. This amplifier circuit, which was initially designed in collaboration with engineers at the Palo Alto Research Center, has undergone several round of optimizations in order to achieve the speed and gain necessary for breast CT and radiotherapy applications. Ideally, further optimization to increase the speed even more and minimize noise would be ideal, but such optimizations are becoming computationally expensive.
One possible approach to realize these improvements without incurring excessive computation time is to borrow machine learning (ML) techniques. An initial pilot study is currently underway to apply low-discrepancy sampling techniques to try to identify a "better" circuit design. Based on the results of this pilot study, future work could employ Bayesian optimization that treats the amplifier circuit as a black-box problem.
Improving mercuric iodide converters with Frisch grid structures
Mercuric iodide (HgI2) is an x-ray converter material with a low effective ionization energy (the average amount of x-ray energy required to produce a pair of electron-hole charges). In addition, it can be manufactured in large area suitable for human-sized x-ray imagers. Both properties are very favorable for medical x-ray imaging. However, HgI2 employed for x-ray imaging exhibits a high degree of image lag (where the signal from a given x-ray exposure takes a long time to exit the converter material). One hypothesis for this lag is due to the slow hole charge carriers.
One proposed method for suppressing these hole carriers is to introduce a Frisch grid structure into the HgI2. This Frisch grid would amplify the signal of the electron charges while decreasing the signal of the hole carriers. Finite element analysis simulations are currently being performed to identify the ideal configuration of Frisch grid wires within the HgI2 material and HgI2 prototypes are being manufactured and characterized. (This work is being performed by a PhD student in the lab that I am helping to mentor.)
Empirical noise performance of prototype active pixel arrays employing polycrystalline silicon thin-film transistorsKoniczek M, Antonuk LE, El-Mohri Y, Liang AK, Zhao Q
Medical Physics 47(6), (2020)
Count rate capabilities of polycrystalline silicon photon counting detectors for CBCT applications – A theoretical studyLiang AK, El-Mohri Y, Zhao Q, Koniczek M, Antonuk L
Physics in Medicine and Biology, 65 035009 (2020)
Theoretical investigation of the count rate capabilities of in-pixel amplifiers for photon counting arrays based on polycrystalline silicon TFTsLiang AK, Koniczek M, Antonuk LE, El-Mohri Y, Zhao Q
Medical Physics 45(10), 4418-4429 (2018)
Theoretical investigation of the noise performance of active pixel imaging arrays based on polycrystalline silicon thin film transistorsKoniczek M, Antonuk LE, El-Mohri Y, Liang AK, Zhao Q.
Medical Physics 44(7), 3491-3503 (2017)
Performance of in-pixel circuits for photon counting arrays (PCAs) based on polycrystalline silicon TFTsLiang AK, Koniczek M, Antonuk LE, El-Mohri Y, Zhao Q, Street RA, Lu JP
Physics in Medicine and Biology 61(5), 1968 (2016)
Expression of recombinant MDA-BF-1 with a kinase recognition site and a 7-histidine tag for receptor binding and purificationLiang AK, Liu J, Mao SA, Siu VS, Lee YC, Lin SH
Protein Expression and Purification 44(1), 55-64 (2005)
16-kDa prolactin down-regulates inducible nitric oxide synthase expression through inhibition of the signal transducer and activator of transcription 1/IFN regulatory factor-1 pathwayLee S, Nishino M, Mazumdar T, Garcia GE, Galfione M, Lee FL, Lee CL, Liang AK, Kim J, Feng L
Cancer Research 65(17), 7984-7992 (2005)