Publications

We develop a high-resolution and high-throughput laboratory-based micro–computed laminography system, which is capable of in situ imaging an industry-relevant large pouch cell at a three-dimensional spatial resolution of 0.5 μm and identifying sub-micron features in multiple cell components. The methodology presents an avenue toward a thorough understanding of the correlation among multiscale structures, chemomechanical degradation, and electrochemical behavior of industry-scale battery pouch cells.

We review our recent findings on charge–lattice–morphology–kinetics in LIB cathode materials in SSRL Liu group. With an emphasis on the data-driven approaches for researching battery materials with synchrotron X-ray techniques, we hope that this Account will lead to more endeavors in this research field.

We propose a novel deep defocus deblurring network that leverages the strength and overcomes the shortcoming of light fields. This strategy is proved to be highly effective and able to achieve the state-of-the-art performance both quantitatively and qualitatively on multiple test sets. Extensive ablation studies have been conducted to analyze the effect of each network module to the final performance.

We developed a deep learning-based approach to extract the crack patterns from nanoscale hard X-ray holo-tomography data of a commercial 18650-type battery cathode. Efficient and effective quantification of the damage heterogeneity with automation and statistical significance is demonstrated. The crack characteristics are further associated with the active particles’ packing densities and a potentially viable architectural design is discussed for suppressing the structural degradation in an industry-relevant battery configuration.

We quantified the characteristics of every single NMC76 particle in its size, sphericity, SOC, SOC variation and anisotropic polarization. To this end, our machine-learning-assisted statistical analysis has revealed the key morphological features that are most prone to the influence by the interphase. The comparison suggests that the LiDFP-induced suppression of both the SOC variation and the anisotropic polarization is more pronounced in the particles with smaller volume and with higher sphericity.

We used hard x-ray holotomography to visualize the structure of a battery composite cathode. With the help of morphology-informed neural network, we are able to track the behavior of thousands of individual particles with time and thus determine the relationship between structure and performance as well as the deterioration of the cathode at a size scale that is not generally accessible. We found that damage during cycling is driven not only by each particle but also by its surrounding neighbors, although the contributions shift over time. This work suggests ways to better design electrodes to maximize their performance.

We tackle the challenge of probing the meso-scale heterogeneity and evolution of lattice defects with sensitivity to atomic-scale details by a unique combination of X-ray nanoprobe diffractive imaging and advanced machine learning techniques. These results pave a direct way to the understanding of crystalline battery materials’ response under external stimuli with high fidelity, which provides valuable empirical guidance to defect-engineering strategies for improving the cathode materials against aggressive battery operation.

We discuss the emergence of synergistic machine-and-data intelligence in synchrotron technology, and how it may accelerate scientific discovery. The optimization should be done jointly with hardware and software, and in an end-to-end fashion.

We adopt a holistic experimental approach to systematically elucidate multiphase, multiscale chemomechanical behaviors in LIB cathodes at extreme low temperatures. Our results suggest that, in order to design batteries for use in a wide temperature range, it is critical to develop electrode components that are structurally and morphologically robust when the cell is switched between different temperatures.

We propose a novel convolutional neural network architecture AIFNet for removing spatially-varying defocus blur from a single defocused image. To remedy the lack of real defocused image datasets, we leverage light field synthetic aperture and refocusing techniques to generate a large set of realistic defocused and all-in-focus image pairs depicting a variety of natural scenes for network training.

We investigate the correlation between the local redox reaction and lattice mismatch through a nano-resolution synchrotron spectro-microscopy study of LiNi0.8Co0.1Mn0.1O2 (NCM 811) cathode particles. With assistance from a machine-learning-based data classification method, we identify local regions that demonstrate a strain–redox decoupling effect, which can be attributed to different side reactions. Our results highlight the mesoscale reaction heterogeneity in the battery cathode and suggest that particle structure engineering could be a viable approach to mitigate the chemomechanical degradation of cathode materials.

We propose an effective computational method to retrieve the image details in high-frame-rate imaging, providing both high temporal and vertical resolution. Experiments and a utility assessment including image data from totally 120 subjects on the measurement of central corneal thickness demonstrate a previously-not-recognized pattern of corneal deformation in air puff tests, as well as the statistically significant difference revealed (p < 0.01) between normal and keratoconus subject groups from the super-resolved time series. The proposed approach is preliminarily proven to be a useful tool for a finer characterization of the cornea.

We adapted the Mask-RCNN model for the identification and quantification of over 650 NMC particles automatically in high-resolution hard X-ray nano-tomography. The statistical analysis reveals that the degree of particle detachment is positively correlated with the charging rate and that smaller particles exhibit a higher degree of uncertainty in their detachment from the carbon/binder matrix.

We propose a new iterative SURE-LET deconvolution algorithm with a plug-in BM3D denoiser. The linear combination of several BM3D denoisers with different (but fixed) parameters, which avoids the manual adjustment of a single non-linear parameter; and linear parametrization makes the minimization of Stein’s unbiased risk estimate (SURE) finally boil down to solving a linear system of equations, leading to a very fast and exact optimization during each iteration.

We propose an approach for estimating the spherically aberrated PSF of a microscope, directly from the observed samples. The PSF, expressed as a linear combination of 4 basis functions, is obtained directly from the acquired image by minimizing a novel criterion, which is derived from the noise statistics in the microscope. The principle of our PSF estimation approach is sufficiently flexible to be generalized non-spherical aberrations and other microscope modalities.

We propose a calibration-free method to obtain the PSF directly from the image obtained. Specifically, we first parametrize the spherically aberrated PSF as a linear combination of few basis functions. The coefficients of these basis functions are then obtained iteratively by minimizing a novel criterion, which is derived from the mixed Poisson-Gaussian noise statistics. Experiments demonstrate that the proposed approach results in highly accurate PSF estimations.

We propose a non-iterative image deconvolution algorithm for data corrupted by Poisson or mixed Poisson-Gaussian noise. In contrast to existing approaches, the proposed algorithm merely amounts to solving a linear system of equations, which has a fast and exact solution. Simulation experiments over different types of convolution kernels and various noise levels indicate that the proposed method outperforms the state-of-the-art techniques, in terms of both restoration quality and computational complexity.

By taking Gaussian kernel as an example, we propose an approach to estimate the blur size for photon-limited images. This estimation is based on the minimization of a novel criterion, blur-PURE (Poisson unbiased risk estimate), which makes use of the Poisson noise statistics of the measurement. Experimental results demonstrate the effectiveness of the proposed method in various scenarios.

We extended the PURE-LET deconvolution algorithm to 3D and applied to deconvolution microscopy. The proposed approach is non-iterative and outperforms existing techniques (usually, variants of Richardson-Lucy algorithm) both in terms of computational efficiency and quality. We illustrate its effectiveness on both synthetic and real data.

A realistic and accurately calculated PSF model can significantly improve the performance in deconvolution microscopy and also the localization accuracy in single-molecule microscopy. We propose a fast and accurate approximation to the Gibson-Lanni model by expressing the integral in this model as a linear combination of rescaled Bessel functions, providing an integral-free way for the calculation. Experiments demonstrate that the proposed approach results in significantly smaller computational time compared with the quadrature approach. This approach can also be extended to other microscopy PSF models.

We propose a non-iterative image deconvolution algorithm for data corrupted by Poisson noise. In contrast to existing approaches, the proposed algorithm merely amounts to solving a linear system of equations which has a fast and exact solution. Simulation experiments over various noise levels indicate that the proposed method outperforms current state-of-the-art techniques, in terms of both restoration quality and computational time.

We propose a novel, multi-frame, tracker that exploits this stationary motion. More precisely, we first estimate the stationary motion and then use it to guide the spot tracker. We obtain the stationary motion by adapting a recent optical flow algorithm that relates one image to another locally using an all-pass filter. We perform this operation over all the image frames simultaneously and estimate a single, stationary optical flow.

We describe the rapidly develop force by the instantaneous rate of force development. The proposed indirect method is a promising alternative to assess I-RFD non-invasively. The torque output is linearly related to the muscle architecture dynamics.

We applied ultrasonography (US) to explore the feasibility of estimating triceps surae force output during isometric plantar flexion with spatial resolution from superficial to deeper muscles. The local deformations of US images are extracted to represent the morphological changes during force generation.

We proposed a coarse-to-fine method based on a compressive-tracking algorithm for estimation of MT changes during an example task of isometric knee extension using ultrasound images. The sensitivity and efficiency are evaluated with 1920 US images from quadriceps muscle (QM) in eight subjects. It is demonstrated that the proposed method agrees well with the manual measurement. Meanwhile, it is not only sensitive to relatively small changes of MT but also computationally efficient.

To provide insights into the rules of longitudinal muscle motion, we proposed a novel framework including motion estimation, visualization and quantitative analysis to interpret synchronous activities of collaborating muscles with spatial details. The proposed framework and its associated quantitative measures could potentially be used to complement electromyography (EMG) and torque signals in functional assessment of skeletal muscles.

In this paper, we proposed a method to estimate the overall orientation of muscle fascicles in a region of interest, in order to complete computing the orientation of the other side of the pennation angle, but the side found by RVHT. The measurements for orientations of both fascicles and aponeurosis were conducted in each frame of ultrasound images, and then the dynamic change of pennation angle during muscle contraction was obtained automatically. The method for fascicle orientation estimation was evaluated using synthetic images with different noise levels and later on 500 ultrasound images of human gastrocnemius muscles during isometric plantarflexion.