BZPEERPapers, people, evidence and research paths
Menu

Discover

SearchFind papers, people, topics, institutions and opportunities from one place.GraphSee how papers, authors, topics and evidence connect.

Research tools

HomeContinue from the papers, people and topics that matter now.CollectionsTurn saved papers into reading lists, evidence maps and shareable context.ResearchReconstruct the path that led you from search to paper to question.CollaborationMove from a paper, topic or expert request into a focused thread.PublishCreate full-text research articles with metadata, formulas and versions.

Network

ExpertsFind specialists whose work and reviews explain why they can help.CommunitiesJoin research groups organized around papers, topics and shared evidence.

Opportunities

ListingsBrowse roles, calls and collaborations connected to real research context.

Account

Log inReturn to your saved papers, graph views and active threads.Create accountStart saving papers, building a research trail and joining teams.
Log inCreate account

Source author record

Ran Gu

Ran Gu appears in the imported research catalog. Authorship, coauthor and topic links are available while profile ownership is still unclaimed.

ResearcherUnclaimed source record
math.COComputer Visioneess.IVComputation and LanguageMachine Learningmath.NAmath.OCNumerical Analysis

Catalog footprint

What is connected

16works
8topics
4close collaborators

Actions

Connect this record

Log in to claim
Open graphBrowse works

Research graph

See the researcher in context

Open full explorer

Inspect adjacent papers, topics, institutions and collaborators without losing the researcher page.

Building this map preview

BZPEER is loading the nearby papers, people, topics and institutions for this page.

Published work

16 published item(s)

preprint2026arXiv

Toward Multimodal Conversational AI for Age-Related Macular Degeneration

Despite strong performance of deep learning models in retinal disease detection, most systems produce static predictions without clinical reasoning or interactive explanation. Recent advances in multimodal large language models (MLLMs) integrate diagnostic predictions with clinically meaningful dialogue to support clinical decision-making and patient counseling. In this study, OcularChat, an MLLM, was fine-tuned from Qwen2.5-VL using simulated patient-physician dialogues to diagnose age-related macular degeneration (AMD) through visual question answering on color fundus photographs (CFPs). A total of 705,850 simulated dialogues paired with 46,167 CFPs were generated to train OcularChat to identify key AMD features and produce reasoned predictions. OcularChat demonstrated strong classification performance in AREDS, achieving accuracies of 0.954, 0.849, and 0.678 for the three diagnostic tasks: advanced AMD, pigmentary abnormalities, and drusen size, significantly outperforming existing MLLMs. On AREDS2, OcularChat remained the top-performing method on all tasks. Across three independent ophthalmologist graders, OcularChat achieved higher mean scores than a strong baseline model for advanced AMD (3.503 vs. 2.833), pigmentary abnormalities (3.272 vs. 2.828), drusen size (3.064 vs. 2.433), and overall impression (2.978 vs. 2.464) on a 5-point clinical grading rubric. Beyond strong objective performance in AMD severity classification, OcularChat demonstrated the ability to provide diagnostic reasoning, clinically relevant explanations, and interactive dialogue, with high performance in subjective ophthalmologist evaluation. These findings suggest that MLLMs may enable accurate, interpretable, and clinically useful image-based diagnosis and classification of AMD.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Contrastive Domain Disentanglement for Generalizable Medical Image Segmentation

Efficiently utilizing discriminative features is crucial for convolutional neural networks to achieve remarkable performance in medical image segmentation and is also important for model generalization across multiple domains, where letting model recognize domain-specific and domain-invariant information among multi-site datasets is a reasonable strategy for domain generalization. Unfortunately, most of the recent disentangle networks are not directly adaptable to unseen-domain datasets because of the limitations of offered data distribution. To tackle this deficiency, we propose Contrastive Domain Disentangle (CDD) network for generalizable medical image segmentation. We first introduce a disentangle network to decompose medical images into an anatomical representation factor and a modality representation factor. Then, a style contrastive loss is proposed to encourage the modality representations from the same domain to distribute as close as possible while different domains are estranged from each other. Finally, we propose a domain augmentation strategy that can randomly generate new domains for model generalization training. Experimental results on multi-site fundus image datasets for optic cup and disc segmentation show that the CDD has good model generalization. Our proposed CDD outperforms several state-of-the-art methods in domain generalizable segmentation.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Contrastive Semi-supervised Learning for Domain Adaptive Segmentation Across Similar Anatomical Structures

Convolutional Neural Networks (CNNs) have achieved state-of-the-art performance for medical image segmentation, yet need plenty of manual annotations for training. Semi-Supervised Learning (SSL) methods are promising to reduce the requirement of annotations, but their performance is still limited when the dataset size and the number of annotated images are small. Leveraging existing annotated datasets with similar anatomical structures to assist training has a potential for improving the model's performance. However, it is further challenged by the cross-anatomy domain shift due to the different appearance and even imaging modalities from the target structure. To solve this problem, we propose Contrastive Semi-supervised learning for Cross Anatomy Domain Adaptation (CS-CADA) that adapts a model to segment similar structures in a target domain, which requires only limited annotations in the target domain by leveraging a set of existing annotated images of similar structures in a source domain. We use Domain-Specific Batch Normalization (DSBN) to individually normalize feature maps for the two anatomical domains, and propose a cross-domain contrastive learning strategy to encourage extracting domain invariant features. They are integrated into a Self-Ensembling Mean-Teacher (SE-MT) framework to exploit unlabeled target domain images with a prediction consistency constraint. Extensive experiments show that our CS-CADA is able to solve the challenging cross-anatomy domain shift problem, achieving accurate segmentation of coronary arteries in X-ray images with the help of retinal vessel images and cardiac MR images with the help of fundus images, respectively, given only a small number of annotations in the target domain.

0 citations0 reviewsNo code linkedOpen access
preprint2022arXiv

Learning towards Synchronous Network Memorizability and Generalizability for Continual Segmentation across Multiple Sites

In clinical practice, a segmentation network is often required to continually learn on a sequential data stream from multiple sites rather than a consolidated set, due to the storage cost and privacy restriction. However, during the continual learning process, existing methods are usually restricted in either network memorizability on previous sites or generalizability on unseen sites. This paper aims to tackle the challenging problem of Synchronous Memorizability and Generalizability (SMG) and to simultaneously improve performance on both previous and unseen sites, with a novel proposed SMG-learning framework. First, we propose a Synchronous Gradient Alignment (SGA) objective, which not only promotes the network memorizability by enforcing coordinated optimization for a small exemplar set from previous sites (called replay buffer), but also enhances the generalizability by facilitating site-invariance under simulated domain shift. Second, to simplify the optimization of SGA objective, we design a Dual-Meta algorithm that approximates the SGA objective as dual meta-objectives for optimization without expensive computation overhead. Third, for efficient rehearsal, we configure the replay buffer comprehensively considering additional inter-site diversity to reduce redundancy. Experiments on prostate MRI data sequentially acquired from six institutes demonstrate that our method can simultaneously achieve higher memorizability and generalizability over state-of-the-art methods. Code is available at https://github.com/jingyzhang/SMG-Learning.

0 citations0 reviewsNo code linkedOpen access
preprint2021arXiv

A fast two-stage algorithm for non-negative matrix factorization in streaming data

In this article, we study algorithms for nonnegative matrix factorization (NMF) in various applications involving streaming data. Utilizing the continual nature of the data, we develop a fast two-stage algorithm for highly efficient and accurate NMF. In the first stage, an alternating non-negative least squares (ANLS) framework is used, in combination with active set method with warm-start strategy for the solution of subproblems. In the second stage, an interior point method is adopted to accelerate the local convergence. The convergence of the proposed algorithm is proved. The new algorithm is compared with some existing algorithms in benchmark tests using both real-world data and synthetic data. The results demonstrate the advantage of our algorithm in finding high-precision solutions.

0 citations0 reviewsNo code linkedOpen access
preprint2021arXiv

Automatic Segmentation of Gross Target Volume of Nasopharynx Cancer using Ensemble of Multiscale Deep Neural Networks with Spatial Attention

Radiotherapy is the main treatment modality for nasopharynx cancer. Delineation of Gross Target Volume (GTV) from medical images such as CT and MRI images is a prerequisite for radiotherapy. As manual delineation is time-consuming and laborious, automatic segmentation of GTV has a potential to improve this process. Currently, most of the deep learning-based automatic delineation methods of GTV are mainly performed on medical images like CT images. However, it is challenged by the low contrast between the pathology regions and surrounding soft tissues, small target region, and anisotropic resolution of clinical CT images. To deal with these problems, we propose a 2.5D Convolutional Neural Network (CNN) to handle the difference of inplane and through-plane resolution. Furthermore, we propose a spatial attention module to enable the network to focus on small target, and use channel attention to further improve the segmentation performance. Moreover, we use multi-scale sampling method for training so that the networks can learn features at different scales, which are combined with a multi-model ensemble method to improve the robustness of segmentation results. We also estimate the uncertainty of segmentation results based on our model ensemble, which is of great importance for indicating the reliability of automatic segmentation results for radiotherapy planning.

0 citations0 reviewsNo code linkedOpen access
preprint2021arXiv

Automatic Segmentation of Organs-at-Risk from Head-and-Neck CT using Separable Convolutional Neural Network with Hard-Region-Weighted Loss

Nasopharyngeal Carcinoma (NPC) is a leading form of Head-and-Neck (HAN) cancer in the Arctic, China, Southeast Asia, and the Middle East/North Africa. Accurate segmentation of Organs-at-Risk (OAR) from Computed Tomography (CT) images with uncertainty information is critical for effective planning of radiation therapy for NPC treatment. Despite the stateof-the-art performance achieved by Convolutional Neural Networks (CNNs) for automatic segmentation of OARs, existing methods do not provide uncertainty estimation of the segmentation results for treatment planning, and their accuracy is still limited by several factors, including the low contrast of soft tissues in CT, highly imbalanced sizes of OARs and large inter-slice spacing. To address these problems, we propose a novel framework for accurate OAR segmentation with reliable uncertainty estimation. First, we propose a Segmental Linear Function (SLF) to transform the intensity of CT images to make multiple organs more distinguishable than existing methods based on a simple window width/level that often gives a better visibility of one organ while hiding the others. Second, to deal with the large inter-slice spacing, we introduce a novel 2.5D network (named as 3D-SepNet) specially designed for dealing with clinic HAN CT scans with anisotropic spacing. Thirdly, existing hardness-aware loss function often deal with class-level hardness, but our proposed attention to hard voxels (ATH) uses a voxel-level hardness strategy, which is more suitable to dealing with some hard regions despite that its corresponding class may be easy. Our code is now available at https://github.com/HiLab-git/SepNet.

0 citations0 reviewsNo code linkedOpen access
preprint2015arXiv

Proper connection number of random graphs

A path in an edge-colored graph is called a proper path if no two adjacent edges of the path are colored the same. For a connected graph $G$, the proper connection number $pc(G)$ of $G$ is defined as the minimum number of colors needed to color its edges, so that every pair of distinct vertices of $G$ is connected by at least one proper path in $G$. In this paper, we show that almost all graphs have the proper connection number 2. More precisely, let $G(n,p)$ denote the Erdös-Rényi random graph model, in which each of the $\binom{n}{2}$ pairs of vertices appears as an edge with probability $p$ independent from other pairs. We prove that for sufficiently large $n$, $pc(G(n,p))\le2$ if $p\ge\frac{\log n +α(n)}{n}$, where $α(n)\rightarrow \infty$.

0 citations0 reviewsNo code linkedOpen access
preprint2014arXiv

An extension of Mantel's theorem to random 4-uniform hypergraphs

A sparse version of Mantel's Theorem is that, for sufficiently large $p$, with high probability (w.h.p.), every maximum triangle-free subgraph of $G(n,p)$ is bipartite. DeMarco and Kahn proved this for $p>K \sqrt{\log n/n}$ for some constant $K$, and apart from the value of the constant, this bound is the best possible. Denote by $T_3$ the 3-uniform hypergraph with vertex set $\{a,b,c,d,e\}$ and edge set $\{abc,ade,bde\}$. Frankl and Füredi showed that the maximum 3-uniform hypergraph on $n$ vertices containing no copy of $T_3$ is tripartite for $n> 3000$. For some integer $k$, let $G^k(n,p)$ be the random $k$-uniform hypergraph. Balogh et al. proved that for $p>K \log n/n$ for some constant $K$, every maximum $T_3$-free subhypergraph of $G^3(n,p)$ w.h.p. is tripartite and it does not hold when $p=0.1 \sqrt{\log n}/n$. Denote by $T_4$ the 4-uniform hypergraph with vertex set $\{1,2,3,4,5,6,7\}$ and edge set $\{1234,1235,4567\}$. Pikhurko proved that there is an $n_0$ such that for all $n\ge n_0$, the maximum 4-uniform hypergraph on $n$ vertices containing no copy of $T_4$ is 4-partite. In this paper, we extend this type of extremal problem in random 4-uniform hypergraphs. We show that for some constant $K$ and $p>K \log n/n$, w.h.p. every maximum $T_4$-free subhypergraph of $G^4(n,p)$ is 4-partite.

0 citations0 reviewsNo code linkedOpen access
preprint2014arXiv

Colorful monochromatic connectivity of random graphs

An edge-coloring of a connected graph $G$ is called a {\it monochromatic connection coloring} (MC-coloring, for short), introduced by Caro and Yuster, if there is a monochromatic path joining any two vertices of the graph $G$. Let $mc(G)$ denote the maximum number of colors used in an MC-coloring of a graph $G$. Note that an MC-coloring does not exist if $G$ is not connected, and in this case we simply let $mc(G)=0$. We use $G(n,p)$ to denote the Erdös-Rényi random graph model, in which each of the $\binom{n}{2}$ pairs of vertices appears as an edge with probability $p$ independently from other pairs. For any function $f(n)$ satisfying $1\leq f(n)<\frac{1}{2}n(n-1)$, we show that if $\ell n \log n\leq f(n)<\frac{1}{2}n(n-1)$ where $\ell\in \mathbb{R}^+$, then $p=\frac{f(n)+n\log\log n}{n^2}$ is a sharp threshold function for the property $mc\left(G\left(n,p\right)\right)\ge f(n)$; if $f(n)=o(n\log n)$, then $p=\frac{\log n}{n}$ is a sharp threshold function for the property $mc\left(G\left(n,p\right)\right)\ge f(n)$.

0 citations0 reviewsNo code linkedOpen access
preprint2014arXiv

Randić Incidence Energy of Graphs

Let $G$ be a simple graph with vertex set $V(G) = \{v_1, v_2,\ldots, v_n\}$ and edge set $E(G) = \{e_1, e_2,\ldots, e_m\}$. Similar to the Randić matrix, here we introduce the Randić incidence matrix of a graph $G$, denoted by $I_R(G)$, which is defined as the $n\times m$ matrix whose $(i, j)$-entry is $(d_i)^{-\frac{1}{2}}$ if $v_i$ is incident to $e_j$ and $0$ otherwise. Naturally, the Randić incidence energy $I_RE$ of $G$ is the sum of the singular values of $I_R(G)$. We establish lower and upper bounds for the Randić incidence energy. Graphs for which these bounds are best possible are characterized. Moreover, we investigate the relation between the Randić incidence energy of a graph and that of its subgraphs. Also we give a sharp upper bound for the Randić incidence energy of a bipartite graph and determine the trees with the maximum Randić incidence energy among all $n$-vertex trees. As a result, some results are very different from those for incidence energy.

0 citations0 reviewsNo code linkedOpen access
preprint2014arXiv

Skew Randić Matrix and Skew Randić Energy

Let $G$ be a simple graph with an orientation $σ$, which assigns to each edge a direction so that $G^σ$ becomes a directed graph. $G$ is said to be the underlying graph of the directed graph $G^σ$. In this paper, we define a weighted skew adjacency matrix with Randć weight, the skew Randić matrix ${\bf R_S}(G^σ)$, of $G^σ$ as the real skew symmetric matrix $[(r_s)_{ij}]$ where $(r_s)_{ij} = (d_id_j)^{-\frac{1}{2}}$ and $(r_s)_{ji} = -(d_id_j)^{-\frac{1}{2}}$ if $v_i \rightarrow v_j$ is an arc of $G^σ$, otherwise $(r_s)_{ij} = (r_s)_{ji} = 0$. We derive some properties of the skew Randić energy of an oriented graph. Most properties are similar to those for the skew energy of oriented graphs. But, surprisingly, the extremal oriented graphs with maximum or minimum skew Randić energy are completely different.

0 citations0 reviewsNo code linkedOpen access
preprint2013arXiv

Degree powers in $C_5$-free graphs

Let $G$ be a graph with degree sequence $d_1,d_2,\ldots,d_n$. Given a positive integer $p$, denote by $e_p(G)=\sum_{i=1}^n d_i^p$. Caro and Yuster introduced a Turán-type problem for $e_p(G)$: given an integer $p$, how large can $e_p(G)$ be if $G$ has no subgraph of a particular type. They got some results for the subgraph of particular type to be a clique of order $r+1$ and a cycle of even length, respectively. Denote by $ex_p(n,H)$ the maximum value of $e_p(G)$ taken over all graphs with $n$ vertices that do not contain $H$ as a subgraph. Clearly, $ex_1(n,H)=2ex(n,H)$, where $ex(n,H)$ denotes the classical Turán number. In this paper, we consider $ex_p(n, C_5)$ and prove that for any positive integer $p$ and sufficiently large $n$, there exists a constant $c=c(p)$ such that the following holds: if $ex_p(n, C_5)=e_p(G)$ for some $C_5$-free graph $G$ of order $n$, then $G$ is a complete bipartite graph having one vertex class of size $cn+o(n)$ and the other $(1-c)n+o(n)$.

0 citations0 reviewsNo code linkedOpen access
preprint2013arXiv

Hypergraph Turán numbers of vertex disjoint cycles

The Turán number of a $k$-uniform hypergraph $H$, denoted by $e{x_k}\left({n;H} \right)$, is the maximum number of edges in any $k$-uniform hypergraph $F$ on $n$ vertices which does not contain $H$ as a subgraph. Let $\mathcal{C}_{\ell}^{\left(k \right)}$ denote the family of all $k$-uniform minimal cycles of length $\ell$, $\mathcal{S}(\ell_1,\ldots,\ell_r)$ denote the family of hypergraphs consisting of unions of $r$ vertex disjoint minimal cycles of length $\ell_1,\ldots,\ell_r$, respectively, and $\mathbb{C}_{\ell}^{\left(k \right)}$ denote a $k$-uniform linear cycle of length $\ell$. We determine precisely $e{x_k}\left({n;\mathcal{S}(\ell_1,\ldots,\ell_r)} \right)$ and $e{x_k}\left({n;\mathbb{C}_{\ell_1}^{\left(k \right)}, \ldots, \mathbb{C}_{\ell_r}^{\left(k \right)}} \right)$ for sufficiently large $n$. The results extend recent results of Füredi and Jiang who determined the Turán numbers for single $k$-uniform minimal cycles and linear cycles.

0 citations0 reviewsNo code linkedOpen access
preprint2013arXiv

Some Motzkin-Straus type results for non-uniform hypergraphs

A remarkable connection between the order of a maximum clique and the Lagrangian of a graph was established by Motzkin and Straus in 1965. This connection and its extensions were applied in Turán problems of graphs and uniform hypergraphs. Very recently, the study of Turán densities of non-uniform hypergraphs has been motivated by extremal poset problems. In this paper, we give some Motzkin-Straus type results for non-uniform hypergraphs.

0 citations0 reviewsNo code linkedOpen access
preprint2013arXiv

The generalized 3-connectivity of random graphs

The generalized connectivity of a graph $G$ was introduced by Chartrand et al. Let $S$ be a nonempty set of vertices of $G$, and $κ(S)$ be defined as the largest number of internally disjoint trees $T_1, T_2, \cdots, T_k$ connecting $S$ in $G$. Then for an integer $r$ with $2 \leq r \leq n$, the {\it generalized $r$-connectivity} $κ_r(G)$ of $G$ is the minimum $κ(S)$ where $S$ runs over all the $r$-subsets of the vertex set of $G$. Obviously, $κ_2(G)=κ(G)$, is the vertex connectivity of $G$, and hence the generalized connectivity is a natural generalization of the vertex connectivity. Similarly, let $λ(S)$ denote the largest number $k$ of pairwise edge-disjoint trees $T_1, T_2, \ldots, T_k$ connecting $S$ in $G$. Then the {\it generalized $r$-edge-connectivity} $λ_r(G)$ of $G$ is defined as the minimum $λ(S)$ where $S$ runs over all the $r$-subsets of the vertex set of $G$. Obviously, $λ_2(G) = λ(G)$. In this paper, we study the generalized 3-connectivity of random graphs and prove that for every fixed integer $k\geq 1$, $$p=\frac{\log n+(k+1)\log \log n -\log \log \log n}{n}$$ is a sharp threshold function for the property $κ_3(G(n, p)) \geq k$, which could be seen as a counterpart of Bollobás and Thomason's result for vertex connectivity. Moreover, we obtain that $δ(G(n,p)) - 1 = λ(G(n,p)) - 1 = κ(G(n,p)) - 1 \le {κ_3}(G(n,p)) \le {λ_3}(G(n,p)) \le κ(G(n,p)) = λ(G(n,p)) = δ(G(n,p))$ almost surely holds, which could be seen as a counterpart of Ivchenko's result.

0 citations0 reviewsNo code linkedOpen access