arXiv subjects

Ye Li

Publications and source records attributed to Ye Li.

At least 181 records · Page 10Linked to original sources

arXiv2022

A Noise-level-aware Framework for PET Image Denoising

In PET, the amount of relative (signal-dependent) noise present in different body regions can be significantly different and is inherently related to the number of counts present in that region. The number of counts in a region depends, in principle and among other factors, on the total administered activity, scanner sensitivity, image acquisition duration, radiopharmaceutical tracer uptake in the region, and patient local body morphometry surrounding the region. In theory, less amount of denoising operations is needed to denoise a high-count (low relative noise) image than images a low-count (high relative noise) image, and vice versa. The current deep-learning-based methods for PET image denoising are predominantly trained on image appearance only and have no special treatment for images of different noise levels. Our hypothesis is that by explicitly providing the local relative noise level of the input image to a deep convolutional neural network (DCNN), the DCNN can outperform itself trained on image appearance only. To this end, we propose a noise-level-aware framework denoising framework that allows embedding of local noise level into a DCNN. The proposed is trained and tested on 30 and 15 patient PET images acquired on a GE Discovery MI PET/CT system. Our experiments showed that the increases in both PSNR and SSIM from our backbone network with relative noise level embedding (NLE) versus the same network without NLE were statistically significant with p<0.001, and the proposed method significantly outperformed a strong baseline method by a large margin.

eess.IV↗

arXiv2021

Hankel Determinants of Certain Sequences Of Bernoulli Polynomials: A Direct Proof of an Inverse Matrix Entry from Statistics

We calculate the Hankel determinants of sequences of Bernoulli polynomials. This corresponding Hankel matrix comes from statistically estimating the variance in nonparametric regression. Besides its entries' natural and deep connection with Bernoulli polynomials, a special case of the matrix can be constructed from a corresponding Vandermonde matrix. As a result, instead of asymptotic analysis, we give a direct proof of calculating an entry of its inverse.

math.NT↗

arXiv2021

Predictions for diphoton production at the LHC through NNLO in QCD

In this paper we present a next-to-next-to-leading order (NNLO) calculation of the process $pp\rightarrow γγ$ that we have implemented into the parton level Monte Carlo code MCFM. We do not find agreement with the previous calculation of this process in the literature. In addition to the $\mathcal{O}(α_s^2)$ corrections present at NNLO, we include some effects arising at $\mathcal{O}(α_s^3)$, namely those associated with gluon-initiated closed fermion loops. We investigate the role of this process in the context of studies of QCD at colliders and as a background for searches for new physics, paying particular attention to the diphoton invariant mass spectrum. We demonstrate that the NNLO QCD prediction for the shape of this spectrum agrees well with functional forms used in recent data-driven fits.

hep-ph↗

arXiv2021

Revisiting the luminosity and redshift distributions of long gamma-ray bursts

In this work, we update and enlarge the long gamma-ray burst (GRB) sample detected by the {\it Swift} satellite. Given the incomplete sampling of the faint bursts and the low completeness in redshift measurement, we carefully select a subsample of bright {\it Swift} bursts to revisit the GRB luminosity function (LF) and redshift distribution by taking into account the probability of redshift measurement. Here we also explore two general expressions for the GRB LF, i.e., a broken power-law LF and a triple power-law LF. Our results suggest that a strong redshift evolution in luminosity (with an evolution index of $δ=1.92^{+0.25}_{-0.37}$) or in density ($δ=1.26^{+0.33}_{-0.34}$) is required in order to well account for the observations, independent of the assumed expression of the GRB LF. However, in a one-on-one comparison using the Akaike information criterion, the best-fitting evolution model involving the triple power-law LF is statistically preferred over the best-fitting one involving the broken power-law LF with a relative probability of $\sim94.3$\% versus $\sim5.7$\%. Extrapolating our fitting results to the flux limit of the whole {\it Swift} sample, and considering the trigger probability of {\it Swift}/Burst Alert Telescope in detail, we find that the expectations from our evolution models provide a good representation of the observed distributions of the whole sample without the need for any adjustment of the model free parameters. This further confirms the reliability of our analysis results.

astro-ph.HE↗

arXiv2021

Searching for radio pulsation from SGR 1935+2154 with the Parkes Ultra-Wideband Low receiver

Magnetars have been proposed to be the origin of FRBs soon after its initial discovery. The detection of the first Galactic FRB 20200428 from SGR 1935+2154 has made this hypothesis more convincing. In October 2020, this source was supposed to be in an extremely active state again. We then carried out a 1.6-hours follow-up observation of SGR 1935+2154 using the new ultra-wideband low (UWL) receiver of the Parkes 64\,m radio telescope covering a frequency range of 704$-$4032 MHz. However, no convincing signal was detected in either of our single pulse or periodicity searches. We obtained a limit on the flux density of periodic signal of $\rm 3.6\,μJy$ using the full 3.3GHz bandwidth data sets, which is the strictest limit for that of SGR 1935+2154. Our full bandwidth limit on the single pulses fluence is 35mJy ms, which is well below the brightest single pulses detected by the FAST radio telescope just two before our observation. Assuming that SGR 1935+2154 is active during our observation, our results suggest that its radio bursts are either intrinsically narrowband or show a steep spectrum.

astro-ph.HE↗

arXiv2021

Cascaded Context Enhancement Network for Automatic Skin Lesion Segmentation

Skin lesion segmentation is an important step for automatic melanoma diagnosis. Due to the non-negligible diversity of lesions from different patients, extracting powerful context for fine-grained semantic segmentation is still challenging today. Although the deep convolutional neural network (CNNs) have made significant improvements on skin lesion segmentation, they often fail to reserve the spatial details and long-range dependencies context due to consecutive convolution striding and pooling operations inside CNNs. In this paper, we formulate a cascaded context enhancement neural network for automatic skin lesion segmentation. A new cascaded context aggregation (CCA) module with a gate-based information integration approach is proposed to sequentially and selectively aggregate original image and multi-level features from the encoder sub-network. The generated context is further utilized to guide discriminative features extraction by the designed context-guided local affinity (CGL) module. Furthermore, an auxiliary loss is added to the CCA module for refining the prediction. In our work, we evaluate our approach on four public skin dermoscopy image datasets. The proposed method achieves the Jaccard Index (JA) of 87.1%, 80.3%, 83.4%, and 86.6% on ISIC-2016, ISIC-2017, ISIC-2018, and PH2 datasets, which are higher than other state-of-the-art models respectively.

eess.IV↗

arXiv2021

Learning Fuzzy Clustering for SPECT/CT Segmentation via Convolutional Neural Networks

Quantitative bone single-photon emission computed tomography (QBSPECT) has the potential to provide a better quantitative assessment of bone metastasis than planar bone scintigraphy due to its ability to better quantify activity in overlapping structures. An important element of assessing response of bone metastasis is accurate image segmentation. However, limited by the properties of QBSPECT images, the segmentation of anatomical regions-of-interests (ROIs) still relies heavily on the manual delineation by experts. This work proposes a fast and robust automated segmentation method for partitioning a QBSPECT image into lesion, bone, and background. We present a new unsupervised segmentation loss function and its semi- and supervised variants for training a convolutional neural network (ConvNet). The loss functions were developed based on the objective function of the classical Fuzzy C-means (FCM) algorithm. We conducted a comprehensive study to compare our proposed methods with ConvNets trained using supervised loss functions and conventional clustering methods. The Dice similarity coefficient (DSC) and several other metrics were used as figures of merit as applied to the task of delineating lesion and bone in both simulated and clinical SPECT/CT images. We experimentally demonstrated that the proposed methods yielded good segmentation results on a clinical dataset even though the training was done using realistic simulated images. A ConvNet-based image segmentation method that uses novel loss functions was developed and evaluated. The method can operate in unsupervised, semi-supervised, or fully-supervised modes depending on the availability of annotated training data. The results demonstrated that the proposed method provides fast and robust lesion and bone segmentation for QBSPECT/CT. The method can potentially be applied to other medical image segmentation applications.

cs.CV↗

arXiv2021

ViT-V-Net: Vision Transformer for Unsupervised Volumetric Medical Image Registration

In the last decade, convolutional neural networks (ConvNets) have dominated and achieved state-of-the-art performances in a variety of medical imaging applications. However, the performances of ConvNets are still limited by lacking the understanding of long-range spatial relations in an image. The recently proposed Vision Transformer (ViT) for image classification uses a purely self-attention-based model that learns long-range spatial relations to focus on the relevant parts of an image. Nevertheless, ViT emphasizes the low-resolution features because of the consecutive downsamplings, result in a lack of detailed localization information, making it unsuitable for image registration. Recently, several ViT-based image segmentation methods have been combined with ConvNets to improve the recovery of detailed localization information. Inspired by them, we present ViT-V-Net, which bridges ViT and ConvNet to provide volumetric medical image registration. The experimental results presented here demonstrate that the proposed architecture achieves superior performance to several top-performing registration methods.

eess.IV↗

arXiv2020

A comparison between repeating bursts of FRB 121102 and giant pulses from Crab pulsar and its applications

There are some similarities between bursts of repeating fast radio bursts (FRBs) and giant pulses (GPs) of pulsars. To explore possible relations between them, we study the cumulative energy distributions of these two phenomena using the observations of repeating FRB 121102 and the GPs of Crab pulsar. We find that the power-law slope of GPs (with fluence $\geq$130 Jy ms) is $2.85\pm0.10$. The energy distribution of FRB 121102 can be well fitted by a smooth broken power-law function. For the bursts of FRB 121102 above the break energy (1.22 $\times 10^{37}$ erg), the best-fitting slope is $2.90_{-0.44}^{+0.55}$, similar to the index of GPs at the same observing frequency ($\sim$1.4 GHz). We further discuss the physical origin of the repeating FRB 121102 in the framework of the super GPs model. And we find that the super GPs model involving a millisecond pulsar is workable and favored for explaining FRB 121102 despite that the magnetar burst model is more popular.

astro-ph.HE↗

arXiv2020

A Multi-task Joint Framework for Real-time Person Search

Person search generally involves three important parts: person detection, feature extraction and identity comparison. However, person search integrating detection, extraction and comparison has the following drawbacks. Firstly, the accuracy of detection will affect the accuracy of comparison. Secondly, it is difficult to achieve real-time in real-world applications. To solve these problems, we propose a Multi-task Joint Framework for real-time person search (MJF), which optimizes the person detection, feature extraction and identity comparison respectively. For the person detection module, we proposed the YOLOv5-GS model, which is trained with person dataset. It combines the advantages of the Ghostnet and the Squeeze-and-Excitation (SE) block, and improves the speed and accuracy. For the feature extraction module, we design the Model Adaptation Architecture (MAA), which could select different network according to the number of people. It could balance the relationship between accuracy and speed. For identity comparison, we propose a Three Dimension (3D) Pooled Table and a matching strategy to improve identification accuracy. On the condition of 1920*1080 resolution video and 500 IDs table, the identification rate (IR) and frames per second (FPS) achieved by our method could reach 93.6% and 25.7,

cs.CV↗

arXiv2020

On the energy and redshift distributions of fast radio bursts

Fast radio bursts (FRBs) are millisecond-duration radio transients from cosmological distances. Their isotropic energies follow a power-law distribution with a possible exponential cutoff, but their intrinsic redshift distribution, which contains information about the FRB sources, is not well understood. We attempt to constrain both distributions by means of Monte Carlo simulations and comparing the simulations results with the available FRB specific fluence distribution, dispersion measure (DM) distribution, and the estimated energy distribution data. Two redshift distribution models, one tracking the star formation history of the universe and another tracking compact binary mergers, are tested. For the latter model, we consider three merger delay timescale distribution (Gaussian, log-normal, and power law) models. Two FRB samples detected by Parkes and the Australian Square Kilometre Array Pathfinder (ASKAP), respectively, are used to confront the simulation results. We confirm the $\sim -1.8$ power law index for the energy distribution but the exponential cutoff energy of the distribution, if any, is unconstrained. For the best energy distribution model, none of the redshift distributions we considered are rejected by the data. A future, larger, uniform FRB sample (such as the one collected by the Canadian Hydrogen Intensity Mapping Experiment, CHIME) can provide better constraints on the intrinsic FRB redshift distribution using the methodology presented in this paper.

astro-ph.HE↗

arXiv2020

Discriminatively Constrained Semi-supervised Multi-view Nonnegative Matrix Factorization with Graph Regularization

In recent years, semi-supervised multi-view nonnegative matrix factorization (MVNMF) algorithms have achieved promising performances for multi-view clustering. While most of semi-supervised MVNMFs have failed to effectively consider discriminative information among clusters and feature alignment from multiple views simultaneously. In this paper, a novel Discriminatively Constrained Semi-Supervised Multi-View Nonnegative Matrix Factorization (DCS^2MVNMF) is proposed. Specifically, a discriminative weighting matrix is introduced for the auxiliary matrix of each view, which enhances the inter-class distinction. Meanwhile, a new graph regularization is constructed with the label and geometrical information. In addition, we design a new feature scale normalization strategy to align the multiple views and complete the corresponding iterative optimization schemes. Extensive experiments conducted on several real world multi-view datasets have demonstrated the effectiveness of the proposed method.

cs.LG↗

arXiv2020

Approximate Nearest Neighbor Negative Contrastive Learning for Dense Text Retrieval

Conducting text retrieval in a dense learned representation space has many intriguing advantages over sparse retrieval. Yet the effectiveness of dense retrieval (DR) often requires combination with sparse retrieval. In this paper, we identify that the main bottleneck is in the training mechanisms, where the negative instances used in training are not representative of the irrelevant documents in testing. This paper presents Approximate nearest neighbor Negative Contrastive Estimation (ANCE), a training mechanism that constructs negatives from an Approximate Nearest Neighbor (ANN) index of the corpus, which is parallelly updated with the learning process to select more realistic negative training instances. This fundamentally resolves the discrepancy between the data distribution used in the training and testing of DR. In our experiments, ANCE boosts the BERT-Siamese DR model to outperform all competitive dense and sparse retrieval baselines. It nearly matches the accuracy of sparse-retrieval-and-BERT-reranking using dot-product in the ANCE-learned representation space and provides almost 100x speed-up.

cs.IR↗

arXiv2020

GRB 200415A: A Short Gamma-Ray Burst from a Magnetar Giant Flare?

The giant flares of soft gamma-ray repeaters (SGRs) have long been proposed to contribute to at least a subsample of the observed short gamma-ray bursts (GRBs). In this paper, we perform a comprehensive analysis of the high-energy data of the recent bright short GRB 200415A, which was located close to the Sculptor galaxy. Our results suggest that a magnetar giant flare provides the most natural explanation for most observational properties of GRB 200415A, including its location, temporal and spectral features, energy, statistical correlations, and high-energy emissions. On the other hand, the compact star merger GRB model is found to have difficulty reproducing such an event in a nearby distance. Future detections and follow-up observations of similar events are essential to firmly establish the connection between SGR giant flares and a subsample of nearby short GRBs.

astro-ph.HE↗

arXiv2020

Quality-aware semi-supervised learning for CMR segmentation

One of the challenges in developing deep learning algorithms for medical image segmentation is the scarcity of annotated training data. To overcome this limitation, data augmentation and semi-supervised learning (SSL) methods have been developed. However, these methods have limited effectiveness as they either exploit the existing data set only (data augmentation) or risk negative impact by adding poor training examples (SSL). Segmentations are rarely the final product of medical image analysis - they are typically used in downstream tasks to infer higher-order patterns to evaluate diseases. Clinicians take into account a wealth of prior knowledge on biophysics and physiology when evaluating image analysis results. We have used these clinical assessments in previous works to create robust quality-control (QC) classifiers for automated cardiac magnetic resonance (CMR) analysis. In this paper, we propose a novel scheme that uses QC of the downstream task to identify high quality outputs of CMR segmentation networks, that are subsequently utilised for further network training. In essence, this provides quality-aware augmentation of training data in a variant of SSL for segmentation networks (semiQCSeg). We evaluate our approach in two CMR segmentation tasks (aortic and short axis cardiac volume segmentation) using UK Biobank data and two commonly used network architectures (U-net and a Fully Convolutional Network) and compare against supervised and SSL strategies. We show that semiQCSeg improves training of the segmentation networks. It decreases the need for labelled data, while outperforming the other methods in terms of Dice and clinical metrics. SemiQCSeg can be an efficient approach for training segmentation networks for medical image data when labelled datasets are scarce.

eess.IV↗

arXiv2020

A comparative study of host galaxy properties between Fast Radio Bursts and stellar transients

Recent arcsecond localizations of Fast Radio Bursts and identifications of their host galaxies confirmed their extragalactic origin.While FRB 121102 resides in the bright region of a dwarf star forming galaxy, other FRBs reside in more massive galaxies and are related to older stellar populations. We compare the host galaxy properties of {nine} FRBs with those of several types of stellar transients: from young to old population, long duration gamma ray bursts (LGRBs), superluminous supernovae (SLSNe), Type Ib/Ic supernovae (SN Ibc), Type II supernovae (SN II), type Ia supernovae (SN Ia), and short duration gamma ray bursts (SGRBs). We find that as a whole sample, the stellar mass and star formation rate of the FRB host galaxies prefer a medium to old population, and are against a young population similar to LGRBs and SLSNe by a null probability 0.02. Individually, the FRB 121102 host is consistent with that of young population objects; the FRB 180924 environment is similar to that of SGRBs; and the FRB 190523 environment is similar to those of SN Ia. These results are consistent with the magnetar engine model for FRBs, if both magnetars produced from extreme explosions (GRBs/SLSNe) and from regular channels (e.g. those producing Galactic magnetars) can produce FRBs.

astro-ph.HE↗

arXiv2020

A Comparative Study of Long and Short GRBs. II. A Multi-wavelength Method to distinguish Type II (massive star) and Type I (compact star) GRBs

Gamma Ray Burst (GRBs) are empirically classified as long-duration GRBs (LGRBs, $>$ 2s) and short-duration GRBs (SGRBs, $<$ 2s). Physically they can be grouped into two distinct progenitor categories: those originating from collapse of massive stars (also known as Type II) and those related to mergers of compact stars (also known as Type I). Even though most LGRBs are Type II and most SGRBs are Type I, the duration criterion is not always reliable to determine the physical category of a certain GRB. Based on our previous comprehensive study of the multi-wavelength properties of long and short GRBs, here we utilize the Naive Bayes method to physically classify GRBs as Type I and Type II GRBs based on multi-wavelength criteria. It results in 0.5\% training error rate and 1\% test error rate. Moreover, there is a gap [-1.2, -0.16] in the distribution of the posterior Odds, $\log O({\rm II:I})$, the Type II to Type I probability ratio. Therefore, we propose to use ${\cal O} = \log O({\rm II:I})+0.7$ as the parameter to classify GRBs into Type I ($<0$) or Type II ($>0$). The only confirmed Type I GRB, GRB 170817A, has log $O({\rm II:I})=-10$. According to this criterion, the supernova-less long GRBs 060614 and 060505 belong to Type I, and two controversial short GRBs 090426 and 060121 belong to Type II.

astro-ph.HE↗

arXiv2020

Boundary-aware Context Neural Network for Medical Image Segmentation

Medical image segmentation can provide a reliable basis for further clinical analysis and disease diagnosis. The performance of medical image segmentation has been significantly advanced with the convolutional neural networks (CNNs). However, most existing CNNs-based methods often produce unsatisfactory segmentation mask without accurate object boundaries. This is caused by the limited context information and inadequate discriminative feature maps after consecutive pooling and convolution operations. In that the medical image is characterized by the high intra-class variation, inter-class indistinction and noise, extracting powerful context and aggregating discriminative features for fine-grained segmentation are still challenging today. In this paper, we formulate a boundary-aware context neural network (BA-Net) for 2D medical image segmentation to capture richer context and preserve fine spatial information. BA-Net adopts encoder-decoder architecture. In each stage of encoder network, pyramid edge extraction module is proposed for obtaining edge information with multiple granularities firstly. Then we design a mini multi-task learning module for jointly learning to segment object masks and detect lesion boundaries. In particular, a new interactive attention is proposed to bridge two tasks for achieving information complementarity between different tasks, which effectively leverages the boundary information for offering a strong cue to better segmentation prediction. At last, a cross feature fusion module aims to selectively aggregate multi-level features from the whole encoder network. By cascaded three modules, richer context and fine-grain features of each stage are encoded. Extensive experiments on five datasets show that the proposed BA-Net outperforms state-of-the-art approaches.

eess.IV↗