arXiv subjects

Bin Dong

Publications and source records attributed to Bin Dong.

At least 91 records · Page 5Linked to original sources

arXiv2023

Distortion-Corrected Image Reconstruction with Deep Learning on an MRI-Linac

Magnetic resonance imaging (MRI) is increasingly utilized for image-guided radiotherapy due to its outstanding soft-tissue contrast and lack of ionizing radiation. However, geometric distortions caused by gradient nonlinearity (GNL) limit anatomical accuracy, potentially compromising the quality of tumour treatments. In addition, slow MR acquisition and reconstruction limit the potential for real-time image guidance. Here, we demonstrate a deep learning-based method that rapidly reconstructs distortion-corrected images from raw k-space data for real-time MR-guided radiotherapy applications. We leverage recent advances in interpretable unrolling networks to develop a Distortion-Corrected Reconstruction Network (DCReconNet) that applies convolutional neural networks (CNNs) to learn effective regularizations and nonuniform fast Fourier transforms for GNL-encoding. DCReconNet was trained on a public MR brain dataset from eleven healthy volunteers for fully sampled and accelerated techniques including parallel imaging (PI) and compressed sensing (CS). The performance of DCReconNet was tested on phantom and volunteer brain data acquired on a 1.0T MRI-Linac. The DCReconNet, CS- and PI-based reconstructed image quality was measured by structural similarity (SSIM) and root-mean-squared error (RMSE) for numerical comparisons. The computation time for each method was also reported. Phantom and volunteer results demonstrated that DCReconNet better preserves image structure when compared to CS- and PI-based reconstruction methods. DCReconNet resulted in highest SSIM (0.95 median value) and lowest RMSE (<0.04) on simulated brain images with four times acceleration. DCReconNet is over 100-times faster than iterative, regularized reconstruction methods. DCReconNet provides fast and geometrically accurate image reconstruction and has potential for real-time MRI-guided radiotherapy applications.

physics.med-ph↗

arXiv2023

Deep Reinforcement Learning for Beam Angle Optimization of Intensity-Modulated Radiation Therapy

Objective: Intensity-modulated radiation therapy (IMRT) beam angle optimization (BAO) is a challenging combinatorial optimization problem that is NP-hard. In this study, we aim to develop a personalized BAO algorithm for IMRT that improves the quality of the final treatment. Methods: To improve the quality of IMRT treatment planning, we propose a deep reinforcement learning (DRL)-based approach for IMRT BAO. We consider the task as a sequential decision-making problem and formulate it as a Markov Decision Process. To facilitate the training process, a 3D-Unet is designed to predict the dose distribution for the different number of beam angles, ranging from 1 to 9, to simulate the IMRT environment. By leveraging the simulation model, double deep-Q network (DDQN) and proximal policy optimization (PPO) are used to train agents to select the personalized beam angle sequentially within a few seconds. Results: The treatment plans with beam angles selected by DRL outperform those with clinically used evenly distributed beam angles. For DDQN, the overall average improvement of the CIs is 0.027, 0.032, and 0.03 for 5, 7, and 9 beam angles respectively. For PPO, the overall average improvement of CIs is 0.045, 0.051, and 0.025 for 5, 7, and 9 beam angles respectively. Conclusion: The proposed DRL-based beam angle selection strategy can generate personalized beam angles within a few seconds, and the resulting treatment plan is superior to that obtained using evenly distributed angles. Significance: A fast and automated personalized beam angle selection approach is been proposed for IMRT BAO.

physics.med-ph↗

arXiv2023

Prediction of transonic flow over supercritical airfoils using geometric-encoding and deep-learning strategies

The Reynolds-averaged Navier-Stokes equation for compressible flow over supercritical airfoils under various flow conditions must be rapidly and accurately solved to shorten design cycles for such airfoils. Although deep-learning methods can effectively predict flow fields, the accuracy of these predictions near sensitive regions and their generalizability to large-scale datasets in engineering applications must be enhanced. In this study, a modified vision transformer-based encoder-decoder network is designed for the prediction of transonic flow over supercritical airfoils. In addition, four methods are designed to encode the geometric input with various information points and the performances of these methods are compared. The statistical results show that these methods generate accurate predictions over the complete flow field, with a mean absolute error on the order of 1e-4. To increase accuracy near the shock area, multilevel wavelet transformation and gradient distribution losses are introduced into the loss function. This results in the maximum error that is typically observed near the shock area decreasing by 50%. Furthermore, the models are pretrained through transfer learning on large-scale datasets and finetuned on small datasets to improve their generalizability in engineering applications. The results generated by various pretrained models demonstrate that transfer learning yields a comparable accuracy from a reduced training time.

physics.flu-dyn↗

arXiv2023

Solving Boltzmann equation with neural sparse representation

We consider the neural sparse representation to solve Boltzmann equation with BGK and quadratic collision model, where a network-based ansatz that can approximate the distribution function with extremely high efficiency is proposed. Precisely, fully connected neural networks are employed in the time and spatial space so as to avoid the discretization in space and time. The different low-rank representations are utilized in the microscopic velocity for the BGK and quadratic collision model, resulting in a significant reduction in the degree of freedom. We approximate the discrete velocity distribution in the BGK model using the canonical polyadic decomposition. For the quadratic collision model, a data-driven, SVD-based linear basis is built based on the BGK solution. All these will significantly improve the efficiency of the network when solving Boltzmann equation. Moreover, the specially designed adaptive-weight loss function is proposed with the strategies as multi-scale input and Maxwellian splitting applied to further enhance the approximation efficiency and speed up the learning process. Several numerical experiments, including 1D wave and Sod problems and 2D wave problem, demonstrate the effectiveness of these neural sparse representation methods.

math.NA↗

arXiv2023

Diffusion Model for Generative Image Denoising

In supervised learning for image denoising, usually the paired clean images and noisy images are collected or synthesised to train a denoising model. L2 norm loss or other distance functions are used as the objective function for training. It often leads to an over-smooth result with less image details. In this paper, we regard the denoising task as a problem of estimating the posterior distribution of clean images conditioned on noisy images. We apply the idea of diffusion model to realize generative image denoising. According to the noise model in denoising tasks, we redefine the diffusion process such that it is different from the original one. Hence, the sampling of the posterior distribution is a reverse process of dozens of steps from the noisy image. We consider three types of noise model, Gaussian, Gamma and Poisson noise. With the guarantee of theory, we derive a unified strategy for model training. Our method is verified through experiments on three types of noise models and achieves excellent performance.

cs.CV↗

arXiv2022

Meta-Auto-Decoder for Solving Parametric Partial Differential Equations

Many important problems in science and engineering require solving the so-called parametric partial differential equations (PDEs), i.e., PDEs with different physical parameters, boundary conditions, shapes of computation domains, etc. Recently, building learning-based numerical solvers for parametric PDEs has become an emerging new field. One category of methods such as the Deep Galerkin Method (DGM) and Physics-Informed Neural Networks (PINNs) aim to approximate the solution of the PDEs. They are typically unsupervised and mesh-free, but require going through the time-consuming network training process from scratch for each set of parameters of the PDE. Another category of methods such as Fourier Neural Operator (FNO) and Deep Operator Network (DeepONet) try to approximate the solution mapping directly. Being fast with only one forward inference for each PDE parameter without retraining, they often require a large corpus of paired input-output observations drawn from numerical simulations, and most of them need a predefined mesh as well. In this paper, we propose Meta-Auto-Decoder (MAD), a mesh-free and unsupervised deep learning method that enables the pre-trained model to be quickly adapted to equation instances by implicitly encoding (possibly heterogenous) PDE parameters as latent vectors. The proposed method MAD can be interpreted by manifold learning in infinite-dimensional spaces, granting it a geometric insight. Extensive numerical experiments show that the MAD method exhibits faster convergence speed without losing accuracy than other deep learning-based methods. The project page with code is available: https://gitee.com/mindspore/mindscience/tree/master/MindElec/.

cs.LG↗

arXiv2022

A Review on Deep Learning in Medical Image Reconstruction

Medical imaging is crucial in modern clinics to guide the diagnosis and treatment of diseases. Medical image reconstruction is one of the most fundamental and important components of medical imaging, whose major objective is to acquire high-quality medical images for clinical usage at minimal cost and risk to the patients. Mathematical models in medical image reconstruction or, more generally, image restoration in computer vision, have been playing a prominent role. Earlier mathematical models are mostly designed by human knowledge or hypothesis on the image to be reconstructed, and we shall call these models handcrafted models. Later, handcrafted plus data-driven modeling started to emerge which still mostly relies on human designs, while part of the model is learned from the observed data. More recently, as more data and computation resources are made available, deep learning based models (or deep models) pushed data-driven modeling to the extreme where the models are mostly based on learning with minimal human designs. Both handcrafted and data-driven modeling have their own advantages and disadvantages. One of the major research trends in medical imaging is to combine handcrafted modeling with deep modeling so that we can enjoy benefits from both approaches. The major part of this article is to provide a conceptual review of some recent works on deep modeling from the unrolling dynamics viewpoint. This viewpoint stimulates new designs of neural network architectures with inspiration from optimization algorithms and numerical differential equations. Given the popularity of deep modeling, there are still vast remaining challenges in the field, as well as opportunities which we shall discuss at the end of this article.

eess.IV↗

arXiv2022

MOTR: End-to-End Multiple-Object Tracking with Transformer

Temporal modeling of objects is a key challenge in multiple object tracking (MOT). Existing methods track by associating detections through motion-based and appearance-based similarity heuristics. The post-processing nature of association prevents end-to-end exploitation of temporal variations in video sequence. In this paper, we propose MOTR, which extends DETR and introduces track query to model the tracked instances in the entire video. Track query is transferred and updated frame-by-frame to perform iterative prediction over time. We propose tracklet-aware label assignment to train track queries and newborn object queries. We further propose temporal aggregation network and collective average loss to enhance temporal relation modeling. Experimental results on DanceTrack show that MOTR significantly outperforms state-of-the-art method, ByteTrack by 6.5% on HOTA metric. On MOT17, MOTR outperforms our concurrent works, TrackFormer and TransTrack, on association performance. MOTR can serve as a stronger baseline for future research on temporal modeling and Transformer-based trackers. Code is available at https://github.com/megvii-research/MOTR.

cs.CV↗

arXiv2022

Region-Aware Metric Learning for Open World Semantic Segmentation via Meta-Channel Aggregation

As one of the most challenging and practical segmentation tasks, open-world semantic segmentation requires the model to segment the anomaly regions in the images and incrementally learn to segment out-of-distribution (OOD) objects, especially under a few-shot condition. The current state-of-the-art (SOTA) method, Deep Metric Learning Network (DMLNet), relies on pixel-level metric learning, with which the identification of similar regions having different semantics is difficult. Therefore, we propose a method called region-aware metric learning (RAML), which first separates the regions of the images and generates region-aware features for further metric learning. RAML improves the integrity of the segmented anomaly regions. Moreover, we propose a novel meta-channel aggregation (MCA) module to further separate anomaly regions, forming high-quality sub-region candidates and thereby improving the model performance for OOD objects. To evaluate the proposed RAML, we have conducted extensive experiments and ablation studies on Lost And Found and Road Anomaly datasets for anomaly segmentation and the CityScapes dataset for incremental few-shot learning. The results show that the proposed RAML achieves SOTA performance in both stages of open world segmentation. Our code and appendix are available at https://github.com/czifan/RAML.

cs.CV↗

arXiv2022

A scalable deep learning approach for solving high-dimensional dynamic optimal transport

The dynamic formulation of optimal transport has attracted growing interests in scientific computing and machine learning, and its computation requires to solve a PDE-constrained optimization problem. The classical Eulerian discretization based approaches suffer from the curse of dimensionality, which arises from the approximation of high-dimensional velocity field. In this work, we propose a deep learning based method to solve the dynamic optimal transport in high dimensional space. Our method contains three main ingredients: a carefully designed representation of the velocity field, the discretization of the PDE constraint along the characteristics, and the computation of high dimensional integral by Monte Carlo method in each time step. Specifically, in the representation of the velocity field, we apply the classical nodal basis function in time and the deep neural networks in space domain with the H1-norm regularization. This technique promotes the regularity of the velocity field in both time and space such that the discretization along the characteristic remains to be stable during the training process. Extensive numerical examples have been conducted to test the proposed method. Compared to other solvers of optimal transport, our method could give more accurate results in high dimensional cases and has very good scalability with respect to dimension. Finally, we extend our method to more complicated cases such as crowd motion problem.

cs.LG↗

arXiv2022

Learning invariance preserving moment closure model for Boltzmann-BGK equation

As one of the main governing equations in kinetic theory, the Boltzmann equation is widely utilized in aerospace, microscopic flow, etc. Its high-resolution simulation is crucial in these related areas. However, due to the high dimensionality of the Boltzmann equation, high-resolution simulations are often difficult to achieve numerically. The moment method which was first proposed by Grad is among the popular numerical methods to achieve efficient high-resolution simulations. We can derive the governing equations in the moment method by taking moments on both sides of the Boltzmann equation, which effectively reduces the dimensionality of the problem. However, one of the main challenges is that it leads to an unclosed moment system, and closure is needed to obtain a closed moment system. It is truly an art in designing closures for moment systems and has been a significant research field in kinetic theory. Other than the traditional human designs of closures, the machine learning-based approach has attracted much attention lately in Han et al. and Huang et al. In this work, we propose a machine learning-based method to derive a moment closure model for the Boltzmann-BGK equation. In particular, the closure relation is approximated by a carefully designed deep neural network that possesses desirable physical invariances, i.e., the Galilean invariance, reflecting invariance, and scaling invariance, inherited from the original Boltzmann-BGK equation and playing an important role in the correct simulation of the Boltzmann equation. Numerical simulations on the 1D-1D examples including the smooth and discontinuous initial condition problems, Sod shock tube problem, the shock structure problems, and the 1D-3D examples including the smooth and discontinuous problems demonstrate satisfactory numerical performances of the proposed invariance preserving neural closure method.

math.NA↗

arXiv2022

A Note on Machine Learning Approach for Computational Imaging

Computational imaging has been playing a vital role in the development of natural sciences. Advances in sensory, information, and computer technologies have further extended the scope of influence of imaging, making digital images an essential component of our daily lives. For the past three decades, we have witnessed phenomenal developments of mathematical and machine learning methods in computational imaging. In this note, we will review some of the recent developments of the machine learning approach for computational imaging and discuss its differences and relations to the mathematical approach. We will demonstrate how we may combine the wisdom from both approaches, discuss the merits and potentials of such a combination and present some of the new computational and theoretical challenges it brings about.

eess.IV↗

arXiv2022

Trained Model in Supervised Deep Learning is a Conditional Risk Minimizer

We proved that a trained model in supervised deep learning minimizes the conditional risk for each input (Theorem 2.1). This property provided insights into the behavior of trained models and established a connection between supervised and unsupervised learning in some cases. In addition, when the labels are intractable but can be written as a conditional risk minimizer, we proved an equivalent form of the original supervised learning problem with accessible labels (Theorem 2.2). We demonstrated that many existing works, such as Noise2Score, Noise2Noise and score function estimation can be explained by our theorem. Moreover, we derived a property of classification problem with noisy labels using Theorem 2.1 and validated it using MNIST dataset. Furthermore, We proposed a method to estimate uncertainty in image super-resolution based on Theorem 2.2 and validated it using ImageNet dataset. Our code is available on github.

cs.LG↗

arXiv2021

Layer-Parallel Training of Residual Networks with Auxiliary-Variable Networks

Gradient-based methods for the distributed training of residual networks (ResNets) typically require a forward pass of the input data, followed by back-propagating the error gradient to update model parameters, which becomes time-consuming as the network goes deeper. To break the algorithmic locking and exploit synchronous module parallelism in both the forward and backward modes, auxiliary-variable methods have attracted much interest lately but suffer from significant communication overhead and lack of data augmentation. In this work, a novel joint learning framework for training realistic ResNets across multiple compute devices is established by trading off the storage and recomputation of external auxiliary variables. More specifically, the input data of each independent processor is generated from its low-capacity auxiliary network (AuxNet), which permits the use of data augmentation and realizes forward unlocking. The backward passes are then executed in parallel, each with a local loss function that originates from the penalty or augmented Lagrangian (AL) methods. Finally, the proposed AuxNet is employed to reproduce the updated auxiliary variables through an end-to-end training process. We demonstrate the effectiveness of our methods on ResNets and WideResNets across CIFAR-10, CIFAR-100, and ImageNet datasets, achieving speedup over the traditional layer-serial training method while maintaining comparable testing accuracy.

cs.LG↗

arXiv2021

Implicit Feature Refinement for Instance Segmentation

We propose a novel implicit feature refinement module for high-quality instance segmentation. Existing image/video instance segmentation methods rely on explicitly stacked convolutions to refine instance features before the final prediction. In this paper, we first give an empirical comparison of different refinement strategies,which reveals that the widely-used four consecutive convolutions are not necessary. As an alternative, weight-sharing convolution blocks provides competitive performance. When such block is iterated for infinite times, the block output will eventually convergeto an equilibrium state. Based on this observation, the implicit feature refinement (IFR) is developed by constructing an implicit function. The equilibrium state of instance features can be obtained by fixed-point iteration via a simulated infinite-depth network. Our IFR enjoys several advantages: 1) simulates an infinite-depth refinement network while only requiring parameters of single residual block; 2) produces high-level equilibrium instance features of global receptive field; 3) serves as a plug-and-play general module easily extended to most object recognition frameworks. Experiments on the COCO and YouTube-VIS benchmarks show that our IFR achieves improved performance on state-of-the-art image/video instance segmentation frameworks, while reducing the parameter burden (e.g.1% AP improvement on Mask R-CNN with only 30.0% parameters in mask head). Code is made available at https://github.com/lufanma/IFR.git

cs.CV↗

arXiv2021

Improving Generalizability in Limited-Angle CT Reconstruction with Sinogram Extrapolation

Computed tomography (CT) reconstruction from X-ray projections acquired within a limited angle range is challenging, especially when the angle range is extremely small. Both analytical and iterative models need more projections for effective modeling. Deep learning methods have gained prevalence due to their excellent reconstruction performances, but such success is mainly limited within the same dataset and does not generalize across datasets with different distributions. Hereby we propose ExtraPolationNetwork for limited-angle CT reconstruction via the introduction of a sinogram extrapolation module, which is theoretically justified. The module complements extra sinogram information and boots model generalizability. Extensive experimental results show that our reconstruction model achieves state-of-the-art performance on NIH-AAPM dataset, similar to existing approaches. More importantly, we show that using such a sinogram extrapolation module significantly improves the generalization capability of the model on unseen datasets (e.g., COVID-19 and LIDC datasets) when compared to existing approaches.

eess.IV↗

arXiv2021

2D wavelength-polarization dispersive microspectroscope based on a hybrid plasmonic helical nanostructure

Microspectrometer features remarkable portability and integration, and is prospected to provide quick detection tools to daily life and revolutionary techniques to researchers. For the extremely finite footprint, microspectrometer can hardly work to analyze the polarization feature by placing polarizer in the optical path like conventional spectrometers. Here, we demonstrate a novel 2D wavelength-polarization dispersive microspectroscope based on carbon nanocoil with plasmonic Au nanopariticles (Au/CNC) as a dispersive component. Explored by the microspectrum and Fourier-space microscopy, a unique 2D dispersive characteristic of the Au/CNC is revealed. Along the axis of the coil, Au/CNC disperses light as wavelength to bands of different diffraction orders like a grating. Wavelength of the incident light can be obtained from the position of the signal lines in a quite large visible-near-infrared wavelength range with an acceptable resolution. In the direction perpendicular to the axis of the coil, incident light is dispersed as polarization with bright and dark areas. We can distinguish left- and right-circularly-polarized light, and also obtain the polarization orientation of linearly-polarized light. Based on this effect, a wonderful 2D wavelength-polarization microspectrometer can be built. It not only fulfills the wavelength analysis with a tiny dispersive component, but also simultaneously knows the polarization feature of the incident light in one shot. Moreover, this powerful tool can further evolve new revolutionary techniques via integrated with other systems.

physics.optics↗

arXiv2021

Solving Partial Differential Equations with Point Source Based on Physics-Informed Neural Networks

In recent years, deep learning technology has been used to solve partial differential equations (PDEs), among which the physics-informed neural networks (PINNs) emerges to be a promising method for solving both forward and inverse PDE problems. PDEs with a point source that is expressed as a Dirac delta function in the governing equations are mathematical models of many physical processes. However, they cannot be solved directly by conventional PINNs method due to the singularity brought by the Dirac delta function. We propose a universal solution to tackle this problem with three novel techniques. Firstly the Dirac delta function is modeled as a continuous probability density function to eliminate the singularity; secondly a lower bound constrained uncertainty weighting algorithm is proposed to balance the PINNs losses between point source area and other areas; and thirdly a multi-scale deep neural network with periodic activation function is used to improve the accuracy and convergence speed of the PINNs method. We evaluate the proposed method with three representative PDEs, and the experimental results show that our method outperforms existing deep learning-based methods with respect to the accuracy, the efficiency and the versatility.

cs.LG↗