Colorectal cancer (CRC) is one of the few cancers that have an established dysplasia-carcinoma sequence that benefits from screening. Everyone over 50 years of age in Canada is eligible for CRC screening. About 20% of those people will undergo a biopsy for a pre-neoplastic polyp and, in many cases, multiple polyps. As such, these polyp biopsies make up the bulk of a pathologist's workload. Developing an efficient computational model to help screen these polyp biopsies can improve the pathologist's workflow and help guide their attention to critical areas on the slide. Deep Learning (DL) models face significant challenges in computational pathology (CPath) because of the gigapixel image size of whole-slide images and the scarcity of detailed annotated datasets. It is, therefore, crucial to leverage self-supervised learning (SSL) methods to alleviate the burden and cost of data annotation. However, current research lacks methods to apply SSL frameworks to analyze pathology data effectively. This paper aims to propose an optimized Barlow Twins framework for colorectal polyps screening. We adapt its hyperparameters, augmentation strategy and encoder to the specificity of the pathology data to enhance performance. Additionally, we investigate the best Field of View (FoV) for colorectal polyps screening and propose a new benchmark dataset for CRC screening, made of four types of colorectal polyps and normal tissue, by performing downstream tasking on MHIST and NCT-CRC-7K datasets. Furthermore, we show that the SSL representations are more meaningful and qualitative than the supervised ones and that Barlow Twins benefits from the Swin Transformer when applied to pathology data. Codes are available from https://github.com/AtlasAnalyticsLab/PathBT.

Auditing outsourced data is a fundamental mechanism for detecting data loss or corruption in untrusted storage environments. Existing auditing schemes rely on public-key primitives based on classical hardness assumptions, which are not secure against quantum adversaries. In this paper, we investigate the issue of post-quantum auditing of outsourced data. We design the PQ-Audit auditing scheme, which utilizes a hash-based signature construction. Specifically, PQ-Audit consists of a small number of hash-based signatures and a hypertree to verify the authenticity of the outsourced data. The data owner signs each outsourced data object before storing it, and the storage server saves these signed data objects for subsequent auditing purposes. However, considering the large size of post-quantum signatures, directly auditing all certified data objects is inefficient. To address this issue, PQ-Audit introduces a sampling-based auditing mechanism that draws inspiration from the provable data possession principle. Audits are conducted on a randomly selected subset of outsourced data objects, which reduces the communication and verification costs during the auditing process and enables efficient auditing. The security analysis demonstrates that PQ-Audit can detect data loss or damage with a probability proportional to the sampling parameter, thereby achieving post-quantum security. Additionally, we conducted a large number of experiments, and the results show that PQ-Audit can efficiently audit large-scale outsourced datasets.

Remote Direct Memory Access (RDMA), with its zero-copy capability and microsecond-level communication latency, has become a core communication technology for highperformance distributed in-memory databases and key-value stores. However, under high-concurrency and read-intensive workloads, existing RDMA-based shared lock release mechanisms still heavily rely on Fetch-and-Add (FAA) atomic operations, which cause NIC serialization bottlenecks and positive feedback blocking in the lock release path. These issues significantly degrade system throughput and exacerbate tail latency. To address this critical performance bottleneck, this paper proposes a series of optimization mechanisms for distributed storage systems built on RDMA. First, a quantitative analysis reveals the remote write amplification and path serialization problems that arise during shared lock release. Based on this insight, we design BR_Lock, which employs a fixed-threshold batching strategy to aggregate multiple discrete lock release operations into a single batched submission, effectively mitigating NIC serialization pressure and reducing critical-path overhead. Furthermore, we propose DBR_Lock, a dynamic batching mechanism that adaptively adjusts release thresholds based on access frequency, intelligently distinguishing between “hot” and “cold” locks to improve concurrency and resource utilization under high contention. Experimental results demonstrate that, compared with state-of-the-art RDMA lock management mechanisms, DBR_Lock achieves up to 1.6 × higher throughput, reduces the 99th-percentile tail latency by more than 50%, and decreases the number of FAA atomic operations by approximately 55-70%. These results verify that the proposed mechanisms achieve an effective balance between scalability, latency optimization, and implementation generality, providing a practical and high-performance lock management solution for RDMA-driven distributed systems.

Abstract Link prediction, which aims to infer missing edges or predict future edges based on currently observed graph connections, has emerged as a powerful technique for diverse applications such as recommendation, relation completion, etc. While there is rich literature on link prediction based on node representation learning, direct link embedding is relatively less studied and less understood. One common practice in previous work characterizes a link by manipulate the embeddings of its incident node pairs, which is not capable of capturing effective link features. Moreover, common link prediction methods such as random walks and graph auto-encoder usually rely on full-graph training, suffering from poor scalability and high resource consumption on large-scale graphs. In this paper, we propose Inductive Subgraph Embedding for Link Prediciton (SE4LP) — an end-to-end scalable representation learning framework for link prediction, which utilizes the strong correlation between central links and their neighborhood subgraphs to characterize links. We sample the “link-centric induced subgraphs” as input, with a subgraph-level contrastive discrimination as pretext task, to learn the intrinsic and structural link features via subgraph classification. Extensive experiments on five datasets demonstrate that SE4LP has significant superiority in link prediction in terms of performance and scalability, when compared with state-of-the-art methods. Moreover, further analysis demonstrate that introducing self-supervision in link prediction can significantly reduce the dependence on training data and improve the generalization and scalability of model.

Large language models (LLMs) can understand human instructions, showing their potential for pragmatic applications beyond traditional NLP tasks. However, they still struggle with complex instructions, which can be either complex task descriptions that require multiple tasks and constraints, or complex input that contains long context, noise, heterogeneous information and multi-turn format. Due to these features, LLMs often ignore semantic constraints from task descriptions, generate incorrect formats, violate length or sample count constraints, and be unfaithful to the input text. Existing benchmarks are insufficient to assess LLMs’ ability to understand complex instructions, as they are close-ended and simple. To bridge this gap, we propose CELLO, a benchmark for evaluating LLMs' ability to follow complex instructions systematically. We design eight features for complex instructions and construct a comprehensive evaluation dataset from real-world scenarios. We also establish four criteria and develop corresponding metrics, as current ones are inadequate, biased or too strict and coarse-grained. We compare the performance of representative Chinese-oriented and English-oriented models in following complex instructions through extensive experiments. Resources of CELLO are publicly available at https://github.com/Abbey4799/CELLO.

Self-supervised learning has been gaining attention in the field of sleep stage classification. It learns representations with unlabeled electroencephalography (EEG) signals, which alleviates the cost of labeling for specialists. However, most self-supervised approaches assume only the two augmented views from the same EEG sample is a positive pair, which suffers from the false negative problem. Therefore, we propose a new model named Temporal Augmentation and False Negative Suppression (TA-FNS) to solve the problem. Specifically, it first generates two augmented views for each EEG sample. Then the temporal augmentation module is proposed to learn temporal features during sleep from augmented views. Based on temporal features, intra-view and inter-view sample similarity matrices are calculated. Finally, the false negative suppression module identifies and eliminates potential false negatives according to the consistency between intra-view and inter-view similarity matrices. TA-FNS not only achieves state-of-the-art performances on Sleep-EDF and ISRUC datasets, but also learns semantic representation from EEG of different sleep stages, which demonstrates the effectiveness of it in mitigating the false negative problem.

Sleep stage classification is crucial for sleep quality assessment and disease diagnosis. While supervised methods have demonstrated good performance in sleep stage classification, obtaining large-scale manually labeled datasets remains a challenge. Recently, self-supervised learning has received increasing attention in sleep stage classification. Self-supervised learning uses unlabeled EEG signals to learn representations, reducing the cost of expert labeling. However, the existing self-supervised learning methods often need to manually adjust the data augmentation strategy according to the characteristics of the data, and only learn the representation from the instance level. Therefore, we propose a self-supervised contrastive learning model FSA-TP for sleep stage classification. Firstly, we design a new feature augmentation module for disturbing the temporal features of EEG signals in the feature space to avoid the tedious operation of manually designing data augmentation strategies. Secondly, we propose a temporal prediction module to learn the temporal representation of EEG signals through a cross-view subsequence prediction task. Finally, we improve the quality of negative samples through the negative mixing module. We evaluate the performance of our proposed method on two publicly available sleep datasets. Experimental results show that FSA-TP not only learns meaningful representations but also produces superior performance.

Abstract Low detection sensitivity and high false positives are two main challenges from traditional lung nodule detection model. In order to improve reliability of detection model, we proposed a multi‐scale three‐dimensional convolutional neural network lung nodule detection method which can simultaneously detect lung nodules and reduce false positives. First, a 3D UNet++‐like architecture with an encoding‐decoding structure is regarded as a feature extraction network and combines it with the region proposal network. The multi‐feature fusion method can fully learn features at different levels through the cross‐layer connection between the encoder and the decoder. Second, the layers of the 3D UNet++‐like architecture are connected by residual blocks, which enhance feature reuse and speeds up the convergence speed of the network. Finally, the three input sizes are input into the 3D neural network, and their classification results are merged, thus to find the final nodule determination result. Based on the LUNA16 dataset, the experiment result shows that average sensitivity is 87.3% and be increased 7.8% than the UNet++ network in our model. And if the number of candidate nodules is 48, the sensitivity is as high as 96.2%, it increases 8.1% than the VGG16 network. Obviously, that the improved model can significantly improve detection sensitivity and reduce false positives, which can provide a valuable theoretical reference for clinical medicine.

The lack of well-annotated datasets in computational pathology (CPath) obstructs the application of deep learning techniques for classifying medical images. Many CPath workflows involve transferring learned knowledge between various image domains through transfer learning. Currently, most transfer learning research follows a model-centric approach, tuning network parameters to improve transfer results over few datasets. In this paper, we take a data-centric approach to the transfer learning problem and examine the existence of generalizable knowledge between histopathological datasets. First, we create a standardization workflow for aggregating existing histopathological data. We then measure inter-domain knowledge by training ResNet18 models across multiple histopathological datasets, and cross-transferring between them to determine the quantity and quality of innate shared knowledge. Additionally, we use weight distillation to share knowledge between models without additional training. We find that hard to learn, multi-class datasets benefit most from pretraining, and a two stage learning framework incorporating a large source domain such as ImageNet allows for better utilization of smaller datasets. Furthermore, we find that weight distillation enables models trained on purely histopathological features to outperform models using external natural image data.

Worldwide, 79.6 million people experience glaucoma, which can cause visual field loss to the individuals. Visual field examination plays an important role in the detection of glaucoma. However, current visual field examination approaches have disadvantages, such as requirements for expensive equipment, long testing time, location restrictions, and more. In the present study, we propose an assessment based on saccadic reaction time (SRT) to overcome the issues in the existing approaches. To confirm the effectiveness of our method, we simulated different stages of glaucoma using a visual field defect simulation system. The result of the visual experiment showed that the discrimination method using SRT can distinguish different symptoms with less testing time.

Though outdoor navigation has been used for a long time, indoor navigation has not yet been put into practice. Recent works based on geomagnetic information require a user to follow the exact path of previous users, which limits its applicability. Meanwhile, the optimal navigation path may not be provided given a large number of paths with different sources and destinations. We present MeshMap, an efficient indoor navigation system by merging geomagnetic information of crowdsourcing paths. MeshMap combines geomagnetic information from multiple users' paths to construct a global navigation map. Then, MeshMap supports to find the optimal navigation path on the navigation map. Finally, MeshMap proposes a real-time tracking method to map user's position on the navigation path and provides real-time navigation hints. We implement MeshMap on Android and evaluate its performance in different environments including campus building, parking area and shopping mall. The evaluation results show the effectiveness of MeshMap.

An efficient and secure environment is necessary for data transmission and storage, especially for large‐column multimedia data. In this study, a novel compression–encryption scheme is presented using a fractal dictionary and Julia set. For the compression in this scheme, fractal dictionary encoding not only reduces time consumption, but also gives good quality image reconstruction. For the encryption in the scheme, the key has large key space and high sensitivity, even to tiny perturbation. Besides, the stream cipher encryption and the diffusion process adopted in this study help spread perturbation in the plaintext, achieving high plain sensitivity and giving an effective resistance to chosen‐plaintext attacks.

Compared with Generalized Gaussian distribution (GGD), Cauchy distribution is superior to describe the statistical distribution of the Intra-coded DCT coefficients in H.264/AVC For the bipolar additive watermark in H.264/AVC video stream, a Cauchy distribution based detection algorithm is proposed by ternary hypothesis testing. Experimental results show that the proposed approach can achieve more than 80% on average for the accuracy of watermark detection.

BACKGROUND: Network co-regulated modules are believed to have the functionality of packaging multiple biological entities, and can thus be assumed to coordinate many biological functions in their network neighbouring regions. RESULTS: Here, we weighted edges of a human protein interaction network and a transcriptional regulatory network to construct an integrated network, and introduce a probabilistic model and a bipartite graph framework to exploit human co-regulated modules and uncover their specific features in packaging different biological entities (genes, protein complexes or metabolic pathways). Finally, we identified 96 human co-regulated modules based on this method, and evaluate its effectiveness by comparing it with four other methods. CONCLUSIONS: Dysfunctions in co-regulated interactions often occur in the development of cancer. Therefore, we focussed on an example co-regulated module and found that it could integrate a number of cancer-related genes. This was extended to causal dysfunctions of some complexes maintained by several physically interacting proteins, thus coordinating several metabolic pathways that directly underlie cancer.

Abstract Motivation: Complex diseases are generally thought to be under the influence of one or more mutated risk genes as well as genetic and environmental factors. Many traditional methods have been developed to identify susceptibility genes assuming a single-gene disease model (‘single-locus methods’). Pathway-based approaches, combined with traditional methods, consider the joint effects of genetic factor and biologic network context. With the accumulation of high-throughput SNP datasets and human biologic pathways, it becomes feasible to search for risk pathways associated with complex diseases using bioinformatics methods. By analyzing the contribution of genetic factor and biologic network context in KEGG (Kyoto Encyclopedia of Genes and Genomes) pathways, we proposed an approach to prioritize risk pathways for complex diseases: Prioritizing Risk Pathways fusing SNPs and pathways (PRP). A risk-scoring (RS) measurement was used to prioritize risk biologic pathways. This could help to demonstrate the pathogenesis of complex diseases from a new perspective and provide new hypotheses. We introduced this approach to five complex diseases and found that these five diseases not only share common risk pathways, but also have their specific risk pathways, which is verified by literature retrieval. Availability: Genotype frequencies of five case–control samples were downloaded from the WTCCC online system and the address is https://www.wtccc.org.uk/info/access_to_data_samples.shtml Contact: [email protected]; [email protected] Supplementary information: Supplementary data are available at Bioinformatics online.