Xiang Zhang

Bio: Xiang Zhang is an academic researcher from Baylor College of Medicine. The author has contributed to research in topics: Medicine & Computer science. The author has an hindex of 154, co-authored 1733 publications receiving 117576 citations. Previous affiliations of Xiang Zhang include University of California, Berkeley & University of Texas MD Anderson Cancer Center.

Intensity modulated proton therapy treatment planning using single-field optimization: The impact of monitor unit constraints on plan quality

Abstract: Purpose: To investigate the effect of monitor unit (MU) constraints on the dose distribution created by intensity modulated proton therapy (IMPT) treatment planning using single-field optimization (SFO). Methods: Ninety-four energies between 72.5 and 221.8 MeV are available for scanning beam IMPT delivery at our institution. The minimum and maximum MUs for delivering each pencil beam (spot) are 0.005 and 0.04, respectively. These MU constraints are not considered during optimization by the treatment planning system; spots are converted to deliverable MUs during postprocessing. Treatment plans for delivering uniform doses to rectangular volumes with and without MU constraints were generated for different target doses, spot spacings, spread-out Bragg peak (SOBP) widths, and ranges in a homogeneous phantom. Four prostate cancer patients were planned with and without MU constraints using different spot spacings. Rounding errors were analyzed using an in-house software tool. Results: From the phantom study, the authors have found that both the number of spots that have rounding errors and the magnitude of the distortion of the dose distribution from the ideally optimized distribution increases as the field dose, spot spacing, and range decrease and as the SOBP width increases. From our study of patient plans, it is clear that asmore » the spot spacing decreases the rounding error increases, and the dose coverage of the target volume becomes unacceptable for very small spot spacings. Conclusions: Constraints on deliverable MU for each spot could create a significant distortion from the ideally optimized dose distributions for IMPT fields using SFO. To eliminate this problem, the treatment planning system should incorporate the MU constraints in the optimization process and the delivery system should reliably delivery smaller minimum MUs.« less

Ultrafast modulation of optical metamaterials.

Abstract: We show by pump-probe spectroscopy that the optical response of a fishnet metamaterial can be modulated on the femtosecond time scale. The modulation dynamics is dominated by pump-induced changes in the constituting dielectric medium, but the strength of modulation is dramatically enhanced through the plasmon resonance. The pump-induced spectral responses of the metamaterial provide understanding on how the resonance is modified by pump excitation. Our study suggests that metamaterials can be used as high-speed amplitude/phase modulators with terahertz-bandwidth.

All Optical Interface for Parallel, Remote, and Spatiotemporal Control of Neuronal Activity

Abstract: A key technical barrier to furthering our understanding of complex neural networks has been the lack of tools for the simultaneous spatiotemporal control and detection of activity in a large number of neurons. Here, we report an all-optical system for achieving this kind of parallel and selective control and detection. We do this by delivering spatiotemporally complex optical stimuli through a digital micromirror spatiotemporal light modulator to cells expressing the light-activated ionotropic glutamate receptor (LiGluR), which have been labeled with a calcium dye to provide a fluorescent report of activity. Reliable and accurate spatiotemporal stimulation was obtained on HEK293 cells and cultured rat hippocampal neurons. This technique should be adaptable to in vivo applications and could serve as an optical interface for communicating with complex neural circuits.

Experimental and numerical investigations on microstereolithography of ceramics

Abstract: Microstereolithography (μSL) uses laser light to solidify UV-curable resin mixed with concentrated ceramic powders. During the μSL process, the light scattering from the particle suspension is found to significantly influence the fabrication resolution in both lateral and depth dimensions which are critical for the complex three-dimensional (3D) microfabrication. In this work, we performed Monte Carlo simulations and experimental studies to understand the detailed microscale optical scattering, chemical reaction (polymerization), and their influence on critical fabrication parameters. As a result, it was found that due to the scattering, the fabricated line is wider in width and smaller in depth compared with polymeric fabrication at the same condition. The doping technique that we used substantially reduced the light scattering, which in turn enhanced the fabrication precision and control. In addition, the experimental values of curing depth and radius agreed reasonably well with the theoretical modeling...

Deep Residual Learning for Image Recognition

Abstract: Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers—8× deeper than VGG nets [40] but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions1, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation.

Very Deep Convolutional Networks for Large-Scale Image Recognition

Abstract: In this work we investigate the effect of the convolutional network depth on its accuracy in the large-scale image recognition setting. Our main contribution is a thorough evaluation of networks of increasing depth using an architecture with very small (3x3) convolution filters, which shows that a significant improvement on the prior-art configurations can be achieved by pushing the depth to 16-19 weight layers. These findings were the basis of our ImageNet Challenge 2014 submission, where our team secured the first and the second places in the localisation and classification tracks respectively. We also show that our representations generalise well to other datasets, where they achieve state-of-the-art results. We have made our two best-performing ConvNet models publicly available to facilitate further research on the use of deep visual representations in computer vision.

Deep Residual Learning for Image Recognition

Abstract: Deeper neural networks are more difficult to train. We present a residual learning framework to ease the training of networks that are substantially deeper than those used previously. We explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. We provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset we evaluate residual nets with a depth of up to 152 layers---8x deeper than VGG nets but still having lower complexity. An ensemble of these residual nets achieves 3.57% error on the ImageNet test set. This result won the 1st place on the ILSVRC 2015 classification task. We also present analysis on CIFAR-10 with 100 and 1000 layers. The depth of representations is of central importance for many visual recognition tasks. Solely due to our extremely deep representations, we obtain a 28% relative improvement on the COCO object detection dataset. Deep residual nets are foundations of our submissions to ILSVRC & COCO 2015 competitions, where we also won the 1st places on the tasks of ImageNet detection, ImageNet localization, COCO detection, and COCO segmentation.

Going deeper with convolutions

Abstract: We propose a deep convolutional neural network architecture codenamed Inception that achieves the new state of the art for classification and detection in the ImageNet Large-Scale Visual Recognition Challenge 2014 (ILSVRC14). The main hallmark of this architecture is the improved utilization of the computing resources inside the network. By a carefully crafted design, we increased the depth and width of the network while keeping the computational budget constant. To optimize quality, the architectural decisions were based on the Hebbian principle and the intuition of multi-scale processing. One particular incarnation used in our submission for ILSVRC14 is called GoogLeNet, a 22 layers deep network, the quality of which is assessed in the context of classification and detection.