Yongqi Li, Shen Zhou, Xiaohu Li, Xin Miao, Jintao Wen, Mayi Xu, Jianhao Chen, Birong Pan, Hankun Kang, Yuanyuan Zhu, Ming Zhong, Tieyun Qian. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Synthesis of myocardial infarction enhancement (MIE) images without contrast agents (CAs) has shown great potential to advance myocardial infarction (MI) diagnosis and treatment. It provides results comparable to late gadolinium enhancement (LGE) images, thereby reducing the risks associated with CAs and streamlining clinical workflows. The existing knowledge-and-data-driven approach has made progress in addressing the complex challenges of synthesizing MIE images (i.e., invisible myocardial scars and high inter-individual variability) but still has limitations in the interpretability of kinematic inference, morphological knowledge integration, and kinematic-morphological fusion, thereby reducing the transparency and reliability of the model and causing information loss during synthesis. In this paper, we proposed a knowledge-driven interpretative conditional diffusion model (K-ICDM), which learns kinematic and morphological information from non-enhanced cardiac MR images (CINE sequence and T1 sequence) guided by cardiac knowledge, enabling the synthesis of MIE images. Importantly, our K-ICDM introduces three key innovations that address these limitations, thereby providing interpretability and improving synthesis quality. (1) A novel cardiac causal intervention that generates counterfactual strain to intervene in the inference process from motion maps to abnormal myocardial information, thereby establishing an explicit relationship and providing the clear causal interpretability. (2) A knowledge-driven cognitive combination strategy that utilizes cardiac signal topology knowledge to analyze T1 signal variations, enabling the model to understand how to learn morphological features, thus providing interpretability for morphology capture. (3) An information-specific adaptive fusion strategy that integrates kinematic and morphological information into the conditioning input of the diffusion model based on their specific contributions and adaptively learns their interactions, thereby preserving more detailed information. Experiments on a broad MI dataset with 315 patients show that our K-ICDM achieves state-of-the-art performance in contrast-free MIE image synthesis, improving structural similarity index measure (SSIM) by at least 2.1% over recent methods. These results demonstrate that our method effectively overcomes the limitations of existing methods in capturing the complex relationship between myocardial motion and scar distribution and integrating of static and dynamic sequences, thus enabling the accurate synthesis of subtle scar boundaries.

A key challenge in sensor-based fall prediction is the fact that a fall event can often occur in various configurations of fall poses together with their own spatio-temporal dependencies. This leads us to define a spatio-temporal model to explicitly characterize these internal configurations of poses. In particular, we introduce a graph neural network with spatio-temporal topological structure to encode such latent relations among poses by capturing representative patterns in fall events. Moreover, a human body orientation estimator is devised to capture human low limbs information, and as a result, separate pose dependencies are globally consistent. Empirical evaluations on two benchmark datasets and one in-house dataset suggest our approach significantly outperforms the state-of-the-art methods.

Falls are the leading cause of unintentional human injury , having become a public health event of strong social concern. The fall prediction technology based on wearable inertial sensors is a relatively reliable solution in human activity monitoring, a user scenario with mobility and high information privacy sensitivity, and has the advantages of low cost, small size, and high precision. However, a key challenge in sensor-based fall prediction is the fact that a fall event can often occur in various configurations of fall poses together with their own spatio-temporal dependencies. This leads us to define a spatio-temporal model to explicitly characterize these internal configurations of poses. In particular, we introduce a graph neural network with spatio-temporal topological structure to encode such latent relations among poses by capturing representative patterns in fall events. Moreover, a human body orientation estimator is devised to represent human low limbs information and as a result, separate pose dependencies are globally consistent. Empirical evaluations on two benchmark datasets and one in-house dataset suggest our approach significantly outperforms the state-of-the-art methods.

Data augmentation has been widely introduced into graph-based tasks to improve the generalizability of models. Based on empirical hypothesis, we show that the node in a graph has different importance, the important one is critical for classification task while the unimportant one hurts the performance. However, there are few works in data augmentation addressing the information propagation of nodes with different importance. In this work, we propose a novel graph data augmentation algorithm for graph classification task, called PassAugment, aiming to pass these importance in graph data augmentation. After distinguishing the importance of all nodes in each graph using the saliency map, we design a data augmentation approach including two strategies: (i) randomly adding edges between the important nodes and the other nodes to globally improve the effective information passing, and (ii) randomly removing edges between the unimportant nodes and their neighbors to locally reduce the ineffective information passing. More importantly, our proposed approach as a standalone module can be combined with many GNNs architectures. Experimental results on graph classification task show that our approach consistently improves the accuracy and achieves or closely matches the state-of-the-art performance.

The simulated driving environment provides a low cost and time-saving platform to test the performance of the autonomous vehicle by linkage with existing machine learning approaches. However, most of existing simulated driving environments focus on building flat roads in urban areas. Still, they neglected to endeavour the tough steep, curvy hill roads, such as mountain paths around suburban areas. In this study, by deploying in Unity engine, we developed the first complex mountain driving simulated environment with characterizing continuous curves and up/downhill. Then, two state-of-art reinforcement learning (RL) algorithms are used to train a vehicle agent and test the performance of autonomous vehicles in our developed simulated environment. Also, we set 5 different levels of vehicle's speeds and observe the cumulative rewards during the vehicle agent training. Our demonstration presents the developed environment supports for complex mountain scenario configurations and RL-based autonomous vehicles, and our findings show that the vehicle agent could achieve high cumulative rewards during the training stage, suggesting that our work is a potential new simulation environment for autonomous vehicles research. The demonstration video can be viewed via the link: https://youtu.be/0wSqGeCn-NU.

Semi-supervised learning has achieved many successes in medical image segmentation since it reduces the costs of manually annotating by leveraging abundant unlabeled data. However, these semi-supervised methods lack attention to ambiguous regions (e.g., some edges or corners around the targets), which may lead to meaningless and unreliable guidance. In this paper, we propose a novel semi-supervised segmentation method called Shape-aware Multi-task Learning (SMTL) to address the above issue. Our multi-task framework includes three tasks namely i) the main task for segmentation ii) one auxiliary task for signed distance regression iii) another auxiliary task for contour detection. The multi-task framework jointly predicts probabilistic segmentation maps, signed distance maps (SDMs) and edge maps to collect complementary information in the existing target label. Specifically, these two auxiliary tasks explicitly enforce shape-priors on the segmentation output to generate more accurate masks. Moreover, we design a region-attention-based adversarial learning strategy that enforces the consistency of two auxiliary tasks prediction distributions on the unlabeled and labeled data to make a meaningful and reliable guidance. We evaluate our SMTL on the datasets of the 2018 Atrial Segmentation Challenge and the 2017 Liver Tumor Segmentation Challenge. The results demonstrate that our SMTL achieves improvements and outperforms the state-of-the-art semi-supervised methods.

We describe a new approach for managing aleatoric uncertainty in the Reinforcement Learning (RL) paradigm. Instead of selecting actions according to a single statistic, we propose a distributional method based on the second-order stochastic dominance (SSD) relation. This compares the inherent dispersion of random returns induced by actions, producing a comprehensive evaluation of the environment’s uncertainty. The necessary conditions for SSD require estimators to predict accurate second moments. To accommodate this, we map the distributional RL problem to a Wasserstein gradient flow, treating the distributional Bellman residual as a potential energy functional. We propose a particle-based algorithm for which we prove optimality and convergence. Our experiments characterize the algorithm’s performance and demonstrate how uncertainty and performance are better balanced using an SSD policy than with other risk measures.

OBJECTIVE: The coronavirus disease 2019 (COVID-19) is rapidly spreading inside China and internationally. We aimed to construct a model integrating information from radiomics and deep learning (DL) features to discriminate critical cases from severe cases of COVID-19 using computed tomography (CT) images. METHODS: We retrospectively enrolled 217 patients from three centers in China, including 82 patients with severe disease and 135 with critical disease. Patients were randomly divided into a training cohort (n = 174) and a test cohort (n = 43). We extracted 102 3-dimensional radiomic features from automatically segmented lung volume and selected the significant features. We also developed a 3-dimensional DL network based on center-cropped slices. Using multivariable logistic regression, we then created a merged model based on significant radiomic features and DL scores. We employed the area under the receiver operating characteristic curve (AUC) to evaluate the model's performance. We then conducted cross validation, stratified analysis, survival analysis, and decision curve analysis to evaluate the robustness of our method. RESULTS: The merged model can distinguish critical patients with AUCs of 0.909 (95% confidence interval [CI]: 0.859-0.952) and 0.861 (95% CI: 0.753-0.968) in the training and test cohorts, respectively. Stratified analysis indicated that our model was not affected by sex, age, or chronic disease. Moreover, the results of the merged model showed a strong correlation with patient outcomes. SIGNIFICANCE: A model combining radiomic and DL features of the lung could help distinguish critical cases from severe cases of COVID-19.

This paper studies series and parallel systems with mutually statistically independent and heterogeneous component lifetimes. The system with homogeneous component lifetimes is proved to age relatively faster than that with component lifetimes following the proportional hazard rates (reversed hazard rates) model in terms of the reversed hazard rate (hazard rate). Sufficient conditions for the relative ageing order on systems with component lifetimes belonging to a scale family are developed as well.

A stochastic model is introduced to investigate the security of network systems based on their vulnerability graph abstractions. Instead of doing traditional security analysis, this paper employs the increasing convex order to study the underlying vulnerability graph. The results of this paper can provide an insight on designing a more secure system, as well as an insight on enhancing the security of an existing system.

In testing Cloud environment testing tasks requested by different tenants have many uncertainties. The arriving time, deadline and the number of tasks are unknown in advance. Especially, the relationships between testing tasks and testing environments are very complex. How to efficiently manage these tasks is really a challenging problem. This paper studies the special features of testing tasks and presents a task management framework. We analyze the dependencies and conflicts associated with testing tasks and their related runtime environments, using rule matching mechanism to derive the relationships supported by domain knowledge. Based on these analyses, improved algorithms are introduced to cluster and dynamically schedule testing tasks to minimize the make-span or meet deadlines with the consideration of testing task resource requirements and Cloud resource utilization balance at the same time. A fault tolerance mechanism is built to cope with testing errors, whose results are studied to ameliorate clustering and scheduling algorithms. A suite of experiments compares the effectiveness of the proposed approach with other algorithms.

Traditional security analyses are often geared toward cryptographic primitives or protocols. Although such analyses are necessary, they cannot address a defender's need for insight into which aspects of a networked system having a significant impact on its security, and how to tune its configurations or parameters so as to improve security. This question is known to be notoriously difficult to answer, and the state of the art is that we know little about it. Toward ultimately addressing this question, this paper presents a stochastic model for quantifying security of networked systems. The resulting model captures two aspects of a networked system: 1) the strength of deployed security mechanisms such as intrusion detection systems and 2) the underlying vulnerability graph, which reflects how attacks may proceed. The resulting model brings the following insights: 1) How should a defender “tune” system configurations (e.g., network topology) so as to improve security? 2) How should a defender “tune” system parameters (e.g., by upgrading which security mechanisms) so as to improve security? 3) Under what conditions is the steady-state number of compromised entities of interest below a given threshold with a high probability? Simulation studies are conducted to confirm the analytic results, and to show the tightness of the bounds of certain important metric that cannot be resolved analytically.

How can we protect sensitive data of average users? In this paper, we propose taking advantage of real-life social trust between average users (called “trust-based social networks”) as well as threshold cryptography. This leads to a new type of complex systems, for which we define and characterize the following novel properties: 1) attack-resilience, which captures the consequences of computers getting compromised; 2) security utility of anonymous social networks, which captures the security gained when the underlying social network links are not known to the attacker; 3) security utility of psychological soundness, which captures the security gained when a user keeps a decisive share of its sensitive data; 4) availability, which captures the effect when computers are not always responsive; 5) the trade-off between attack-resilience and availability.

This note studies coherent systems with independent, -identically distributed (i.i.d.) components. Aging properties and stochastic orders are established for the general residual life, and the general inactivity time.

Digital signatures are an important security mechanism, especially when non-repudiation is desired. However, non-repudiation is meaningful only when the private signing keys and functions are adequately protected --- an assumption that is very difficult to accommodate in the real world because computers (and thus cryptographic keys and functions) could be relatively easily compromised. One approach to resolving, or at least alleviating, this problem is to use threshold cryptography. But how should such techniques be employed in the real world? In this paper we propose exploiting social networks whereby average users take advantage of their trusted ones to help secure their cryptographic keys. While the idea is simple from an individual user's perspective, we aim to understand the resulting systems from a whole-system perspective. Specifically, we propose and investigate two measures of the resulting systems: attack-resilience, which captures the security consequences due to the compromise of some computers and thus the compromise of the cryptographic key shares stored on them; availability, which captures the effect when computers are not always responsive (due to the peer-to-peer nature of social networks).

Traditional security analyses are often geared towards cryptographic primitives or protocols. Although such analyses are absolutely necessary, they do not provide much insight for answering an equally important question: what is the security assurance of a physically or logically networked system when we consider it as a whole? This question is known to be notoriously difficult, and the state-of-the-art is that we know very little about it. In this paper, we make a step towards resolving it with a new modeling approach.

Mathematical models have been utilized to help understand the epidemic spreading of malicious codes (e.g., computer virus and worms). However, existing such models are either adapted from the ones developed to capture the epidemic spreading of biologically infectious diseases in homogeneous systems, or suitable only for a very specific class of heterogeneous systems. In this paper we present an attempt at building an analytic model of epidemic spreading of malicious codes in arbitrary heterogeneous systems.

Abstract Boesch et al. conjectured that for any n and m there exists a uniformly optimal (n,m)–graph Gn,m for all terminal reliability, that is, the all‐terminal reliability of Gn,m is at least as large as the all‐terminal reliability of any other graph G with n vertices and m edges, no matter what the probability of an edge being operational is. Although there are counterexamples known when one restricts attention to simple graphs, the conjecture remains open when one allows parallel edges. We consider the analogous problem for strongly connected reliability, that is, the probability that a digraph contains a spanning strongly connected subdigraph, given that each vertex is operational, but arcs are independently operational with probability p. We show that there do indeed exist uniformly optimal digraphs for strongly connected (n,m)–digraphs. We also show that if one restricts attention to simple digraphs (without parallel arcs) then such uniformly optimal digraphs need not exist. © 2006 Wiley Periodicals, Inc. NETWORKS, Vol. 49(2), 145–151 2007

The$k$-out-of-$n$structure is a very popular type of redundancy in fault-tolerant systems. It has been applied in industrial, and military systems. In this paper, we investigate the general residual life (GRL) of a$(n-k+1)$-out-of-$n$system with i.i.d. components, given that the total number of the failures of components is less than$l-1(1leq l≪ kleq n)$at time$tgeq0$. It is shown that the GRL is decreasing in$l$in terms of the likelihood ratio order; Behavior of IFR, and NBU of life distributions are discussed in terms of the monotonicity of GRL. Finally, comparison of the GRL of two$(n-k+1)$-out-of-$n$systems are conducted given that the lifetime of their components are assumed to be ordered in the hazard rate order.

Abstract Given a digraph D, consider the model where each vertex is always operational, but the edges are independently operational with probability p. The strongly connected reliability of D, scRel(D,p), is the probability that the spanning subgraph of D consisting of the operational edges is strongly connected. One can view strongly connected reliability as the probability that any vertex can send information to any other vertex, given that edges fail independently. There are very few classes for which there is an efficient algorithm for calculating the strongly connected reliability. This article presents the fist polynomial time algorithm for computing the strongly connected reliability of complete digraphs, that is, digraphs in which every vertex is joined to every other vertex by exactly one edge (one in each direction). © 2005 Wiley Periodicals, Inc. NETWORKS, Vol. 45(3), 165–168 2005

A new class of life distributions, namely new better than used in the total time on test transform ordering (NBUT), is introduced. The relationship of this class to other classes of life distributions, and closure properties under some reliability operations, are discussed. We provide a simple argument based on stochastic orders that the family of the NBUT distribution class is closed under the formation of series systems in case of independent identically distributed components. Behavior of this class is developed in terms of the monotonicity of the residual life of k-out-of-n systems given the time at which the (n-k)-th failure has occurred. Finally, we discuss testing exponentially against the NBUT aging property.