Segmentation of bone marrow lesions (BML) is vital for quantifying lesion volume, a biomarker associated with cartilage damage and pain in knee osteoarthritis (KOA). Manual segmentation is challenging due to low contrast and variable lesion locations. We propose an automated deep learning method featuring a dual-stream encoder that integrates global image-level and local patch-level features. The study included 300 participants from the Osteoarthritis Initiative (OAI) database, each with approximately 36 intermediate-weighted fat-suppressed (IWFS) magnetic resonance (MR) images. The ground truth masks were manually annotated by trained research staff. With physical batch size 32, the model achieved a 2D Dice similarity coefficient (DSC) of 0.68, 3D DSC of 0.62, Intersection over Union (IoU) of 0.51, precision of 0.76, sensitivity of 0.62, and Pearson’s correlation coefficient (r) of 0.85 between manually labelled and automatically generated volumes. Using an effective batch size of 64 via gradient accumulation, the model achieved 2D DSC of 0.63, 3D DSC of 0.65, IoU of 0.48, precision of 0.75, sensitivity of 0.6, and r of 0.98 for volume correlation. The model outperformed baselines at batch size 32 across almost all evaluated metrics and remained robust at batch size 64, with strong volumetric correlation and improved 3D DSC, IoU, and sensitivity.

Accurate segmentation of polyp tissues in colonoscopic images is crucial for early colorectal cancer detection. Existing CNN-based approaches effectively capture local dependencies but struggle with long-range relations, while transformer-based methods excel in global context modeling yet often overlook fine contextual details. Hybrid CNN–transformer models attempt to combine both, but typically overfit to convolutional features, weakening attention mechanisms. To address these limitations, we propose a Hierarchical Contextual Information Aggregation Network (HCIA) for polyp segmentation. HCIA introduces an Interconnected Attention Module (IAM) that applies global attention to single-level features, enabling comprehensive cross-hierarchy information exchange. In parallel, a Hierarchical Aggregation Module (HAM) fuses adjacent feature levels to enhance local contextual representation. This dual refinement allows HCIA to jointly capture global and local dependencies, yielding more precise tissue boundaries. Extensive experiments across multiple polyp segmentation benchmarks demonstrate that HCIA achieves superior generalization and state-of-the-art accuracy, highlighting its potential for clinical applications.

Autism spectrum disorder (ASD) has been associated with diverse genetic factors and molecular changes. Yet, how pathophysiology emerges during development and contributes to clinical heterogeneity remains unclear. Here we use spatial and single-cell transcriptomics of genetically diverse patient-derived organoids of the prenatal cortex to map the spatial architecture of ASD pathogenesis. We find abnormal partitioning of progenitor and neuron zones alongside local patches of disorganized neurons that vary in distribution among patients. Such spatially mosaic disarray persists into neuronal maturation and is consistent with impaired adhesion between progenitors. Our findings suggest that the spatial landscape of biological processes leading to ASD may be more heterogeneous than previously thought. We propose a model in which different patterns of scattered abnormalities, arising from spatially mosaic pathogenesis during prenatal brain development, contribute to the wide range of symptoms and brain structure variations observed in ASD patients.

Cohomological equation is of special interest because it concerns the study of time change for flows, topological stability and topological conjugacy in dynamical systems, which is also an auxiliary equation to study the problem of linearization. In this paper, we consider a general form of cohomological equation for planar contractions. By using the ideas of invariant manifold and estimations in [W. Zhang and W. Zhang, dollar sign upper C 1 dollar sign $C<^>1$ C 1 linearization for planar contractions, J. Funct. Anal. 260 (2011), 2043-2063.], we present new criteria on eigenvalues of the linear parts for the existence of dollar sign upper C 1 dollar sign $C<^>1$ C 1 solutions in the Poincar & eacute; domain. Our results are a generalization of dollar sign upper C 1 dollar sign $C<^>1$ C 1 linearization for contractions.

Abstract Cohomological equation is of special interest because it concerns the study of time change for flows, topological stability and topological conjugacy in dynamical systems, which is also an auxiliary equation to study the problem of linearization. In this paper, we consider a general form of cohomological equation for planar contractions. By using the ideas of invariant manifold and estimations in [W. Zhang and W. Zhang, dollar sign upper C 1 dollar sign $C^1$ C 1 linearization for planar contractions, J. Funct. Anal. 260 (2011), 2043–2063.], we present new criteria on eigenvalues of the linear parts for the existence of dollar sign upper C 1 dollar sign $C^1$ C 1 solutions in the Poincaré domain. Our results are a generalization of dollar sign upper C 1 dollar sign $C^1$ C 1 linearization for contractions.

Basal roughness is a crucial parameter for quantifying subglacial geomorphological landforms, which offers key insights into glacial geomorphic environments and ice sheet dynamics. Princess Elizabeth Land (PEL) in East Antarctica covers approximately 15% of the Antarctic Ice Sheet. However, to date, understanding the relationship between subglacial geomorphology and ice flow in the PEL has remained limited. In this study, we used airborne ice radar data from the Chinese National Antarctic Research Expedition (CHINARE) during the first five austral seasons and publicly available Antarctica's Gamburtsev Province (AGAP) Project North data to calculate a two-parameter spectral roughness index of the subglacial topography. We analyzed the relationship between the spatial distribution of basal roughness and the speed and direction of ice velocity, while classifying the regional roughness results into four different combinations. We find that the subglacial environment in the PEL is more intricate than the one previously reported. The area near the polar record glacier (PRG) is characterized by locally rough geomorphology but fast ice flow. The beds in the slow ice flow area of PEL are characterized by both rough and flat landforms. Low-lying basins situated in the interior are of considerable interest because they may be characterized by preglacial active erosional landscapes.

We present SPRINT, a novel approach for large-scale, cost-effective synthesis of instruction-tuning datasets, leveraging Program-of-Thoughts (PoT) to enhance mathematical reasoning capabilities. Through the SPRINT framework, we synthesized data from seven high-quality open-source math datasets (including GSM8K, MATH, AQuA), and developed InfinityMATH-a dataset containing over 100,000 samples generated from QA pairs, offering extensive coverage across various mathematical domains. The SPRINT model series, fine-tuned on InfinityMATH using open-source language and code models such as Llama2-7B, Mistral-7B, and CodeLlama7B, achieved remarkable improvements in mathematical reasoning, with performance gains between 184.7% and 514.3%. In zero-shot settings, our SPRINT-CodeLlama-7B model surpassed MAmmoTH-Coder on widelyused benchmarks, including GSM8K (65.80% vs. 56.86%) and MATH (34.06% vs. 29.88%). To assess logical consistency in numerical transformations, we created the GSM8K+ and MATH+ test sets by modifying the numerical values in the original datasets. While traditional models struggled with these alterations, the SPRINT models exhibited superior robustness. The InfinityMATH dataset is publicly available at https://huggingface. co/datasets/BAAI/InfinityMATH.

Autism spectrum disorder (ASD) is characterized by difficulties in social interaction, communication challenges, and repetitive behaviors. Despite extensive research, the molecular mechanisms underlying these neurodevelopmental abnormalities remain elusive. We integrated microscale brain gene expression data with macroscale MRI data from 1829 participants, including individuals with ASD and typically developing controls, from the autism brain imaging data exchange I and II. Using fractal dimension as an index for quantifying cortical complexity, we identified significant regional alterations in ASD, within the left temporoparietal, left peripheral visual, right central visual, left somatomotor (including the insula), and left ventral attention networks. Partial least squares regression analysis revealed gene sets associated with these cortical complexity changes, enriched for biological functions related to synaptic transmission, synaptic plasticity, mitochondrial dysfunction, and chromatin organization. Cell-specific analyses, protein–protein interaction network analysis and gene temporal expression profiling further elucidated the dynamic molecular landscape associated with these alterations. These findings indicate that ASD-related alterations in cortical complexity are closely linked to specific genetic pathways. The combined analysis of neuroimaging and transcriptomic data enhances our understanding of how genetic factors contribute to brain structural changes in ASD.

This paper discusses the admissible consensus tracking problem for nonlinear singular multi-agent systems under sampled-data event-triggered mechanisms, where the nonlinear dynamics are unknown. In the sampled-data framework, both static and dynamic event-triggered mechanisms are designed based on the observers, and the dynamic event-triggered mechanism has superior performance. Moreover, in order to eliminate the effect of bounded consensus caused by the double estimation problem, this study proposes a distributed adaptive event-triggered control protocol. This protocol is designed to establish sufficient conditions for achieving admissible consensus tracking of singular multi-agent systems. Finally, the theoretical results are verified by three numerical examples.

Autism spectrum disorder (ASD) is marked by profound neurobiological heterogeneity, yet it remains unclear whether atypical brain organization reflects a diffuse low-amplitude pattern shared broadly across individuals or distinct spatially specific deviations that vary from person to person. Resolving this question requires methods that move beyond group averages to map individualized cortical atypicality against normative expectations. We built subject-level morphometric similarity networks in which edges quantify multivariate morphometric similarity between cortical regions. We then applied hierarchical Bayesian regression (HBR) normative models to estimate region-wise deviations from age- and sex-adjusted norms while accounting for site variation. Individuals with ASD carried a greater burden of extreme regional deviations, yet those deviations were focal and idiosyncratic rather than uniformly distributed. The spatial pattern of deviation aligned with canonical cortical hierarchies, shifting similarity toward sensory and visual poles and away from association cortex. Moreover, edge-level testing identified a sparse reconfiguration concentrated in occipito-temporal and cross-network connections. Clustering of individual deviation maps identified two robust subgroups that differ in the global sign of deviation and in their cognitive and molecular correlates. These results show that cortical atypicality in ASD is constrained by brain gradients, expressed in subgroup-specific forms, and linked to distinct biological and cognitive axes. ### Competing Interest Statement The authors have declared no competing interest.

The study of organism structural composition, known as anatomy, is essential in comprehending the intricate arrangements of life and plays a crucial role in medical education and practice. It bridges foundational and clinical disciplines, shaping medical education, and practice. With evolving technology, medical education faces new challenges necessitating pedagogical innovations. This article explores the changing landscape of anatomical education, encompassing teaching methods, and curricular shifts. Advancements in information technology and bibliometrics shed light on anatomy's evolution, yet research on anatomy teaching reform (ATR) remains scarce. This study employs advanced analytical tools like CiteSpace and VOSviewer to uncover research hotspots and frontiers in ATR. By scrutinizing focal points and emerging directions in ATR, this research provides insights into the future of pedagogical strategies and clinical research in anatomy.

Steganography is to conceal the presence of secret communication.Steganalyser based on rich models and deep learning achieves state-of-the-art performance.However, due to the difficulty of capturing the unknown distribution of the high-dimensional cover, it is challenging to design a steganographic scheme from the view of traditional and deep learning-based steganography to defeat the steganalyser.In this paper, we propose a scheme to search the steganographic policy from scratch with the help of auxiliary constrained distance measure and the adversary.The auxiliary distance measure is provided via similarity evaluation and cover estimator.The response space of distance measures is constrained to provide better performance.Similarity evaluation establishes the distance measure in cover space.The cover estimator is to predict the potential distribution of the cover.Thus it could provide the distance measure in distribution space.The adversary model is to simulate the role of the steganalyser.The steganographic policy is searched from playing the adversarial game against an adversary with the relaxation of constrained response space of auxiliary distance measure.Ultimately, within the searched steganographic policy, the stego objects could be generated by the STC coder.Cover estimator, adversary, and steganographic policy are parameterized via neural networks.Experiments demonstrate that our scheme could achieve an effective modification policy and has competitive security performance compared with traditional and state-of-the-art deep learning-based steganographic methods.

Due to high transport efficiency, reduced transfer time, and various other advantages, the joint operation of different rail transit systems emerges as the optimal choice for rail transit systems. This article mainly studies the line-planning problem under the line-sharing operation mode between metro and suburban railway. First, a complex multiobjective programming model is established to maximize the net profit of two operating companies and the time savings of passengers. The constraints of this model encompass passenger flow, available vehicles, line carrying capacity, station capacity, cross-line configuration, departure frequency, and variable value range. Second, the linear weighted sum method is introduced to consolidate three objective functions into a single one, while utilizing the improved artificial bee colony (IABC) algorithm to address the line-planning problem. Besides, the traditional artificial bee colony (TABC) algorithm and the simulated annealing (SA) algorithm are provided as comparison groups to solve the same numerical example problem. The results demonstrate significant reductions in travel time by adopting the line-sharing operation mode. In addition, the IABC algorithm exhibits better solution quality and higher efficiency than both the TABC and SA algorithms. The proposed method proves to be valuable in formulating and optimizing the line plan.

Multimodal rail transit systems integration and interconnection can solve frequent transfer problems and better adapt to disequilibrium passenger flow and space. It is an inevitable choice in the development of various rail transit systems. Firstly, this paper proposes a novel train service plan design model in the scenario of multimodal rail transit systems integration and interconnection. Our model takes into account the costs of both passengers and enterprises, and passengers travel time is converted into cost using passengers' nonworking time value coefficient. The model contains some conventional constraints such as passenger flow, station capacity, and line carrying capacity. It also considers whether the transportation capacity of different lines is matched, that is, the constraint of capacity matching degree. Secondly, an improved harmonic search algorithm (IHSA) is designed to solve the problem, and a numerical experiment is used to prove the performance of the proposed method. Our research result shows that the model and algorithm proposed in this paper is effective, which can help overcome the drawbacks of the existing independent operation mode of different rail transit systems. This study is also applicable to the scenario of other kinds of rail transit systems integration and interconnection.

Fractal dimension (FD) is used to quantify brain structural complexity and is more sensitive to morphological variability than other cortical measures. However, the effects of normal aging and sex on FD are not fully understood. In this study, age- and sex-related differences in FD were investigated in a sample of 448 adults age of 19-80 years from a Chinese dataset. The FD was estimated with the surface-based morphometry (SBM) approach, sex differences were analyzed on a vertex level, and correlations between FD and age were examined. Generalized additive models (GAMs) were used to characterize the trajectories of age-related changes in 68 regions based on the Desikan-Killiany atlas. The SBM results showed sex differences in the entire sample and 3 subgroups defined by age. GAM results demonstrated that the FD values of 51 regions were significantly correlated with age. The trajectories of changes can be classified into 4 main patterns. Our results indicate that sex differences in FD are evident across developmental stages. Age-related trajectories in FD are not homogeneous across the cerebral cortex. Our results extend previous findings and provide a foundation for future investigation of the underlying mechanism.

AbstractAs a weak version of embedding flow, the problem of iterative roots is studied extensively in one dimension, especially in monotone case. There are few results in high dimensions because the constructive method dealing with monotone mappings is unavailable. In this paper, by introducing a kind of partial order, we define the monotonicity for two-dimensional mappings and then present some results on the existence of iterative roots for linear mappings, triangle-type mappings, and co-triangle-type mappings, respectively. Our theorems show that even the property of monotonicity for iterative roots of monotone mappings, which is a trivial result in one dimension, does not hold anymore in high dimensions. At the end of this paper, the problem of iterative roots for two well-known planar mappings, that is, Hénon mappings and coupled logistic mappings, are also discussed.

Abstract This paper deals with the finite‐time stability problem of non‐linear singular multi‐agent systems subject to controller gain disturbance via a distributed non‐fragile controller. The existing literature on this problem only applies to normal multi‐agent systems, or ignores the possible change of control gain in the practical application of the controller, or requires that there is no time‐varying delay in the control input. How to tackle the finite‐time stability problem of singular multi‐agent systems by considering input time‐varying delay, Lipschitz non‐linear dynamics and controller gain perturbation simultaneously is an open issue. In order to solve this problem, a distributed non‐fragile controller is proposed for each agent to make the finite‐time stability problem solved for any connected undirected connection topology. With the help of a Linear matrix inequality and a Lyapunov integral appropriate expression, the difficulty caused by the irreversibility of singular system matrices is solved. Then, a sufficient condition to ensure the finite‐time stability of the closed‐loop system is obtained for the singular multi‐agent system including controller disturbance, input–output delay and non‐linear dynamics. Finally, two numerical examples are used to illustrate the effectiveness of the proposed method.

The evolving intercloud enables idle resources to be traded among cloud providers to facilitate utilization optimization and to improve the cost-effectiveness of the service for cloud consumers. However, several challenges are raised for this multi-tier dynamic market, in which cloud providers not only compete for consumer requests but also cooperate with each other. To establish a healthier and more efficient intercloud ecosystem, in this paper a multi-tier agent-based fuzzy constraint-directed negotiation (AFCN) model for a fully distributed negotiation environment without a broker to coordinate the negotiation process is proposed. The novelty of AFCN is the use of a fuzzy membership function to represent imprecise preferences of the agent, which not only reveals the opponent’s behavior preference but can also specify the possibilities prescribing the extent to which the feasible solutions are suitable for the agent’s behavior. Moreover, this information can guide each tier of negotiation to generate a more favorable proposal. Thus, the multi-tier AFCN can improve the negotiation performance and the integrated solution capacity in the intercloud. The experimental results demonstrate that the proposed multi-tier AFCN model outperforms other agent negotiation models and demonstrates the efficiency and scalability of the intercloud in terms of the level of satisfaction, the ratio of successful negotiation, the average revenue of the cloud provider, and the buying price of the unit cloud resource.

With the rapid development of the Internet, the way of education is gradually tending to the network platform, so the intelligent education platform plays an extremely important role in the diversified way of education. It has always been an important issue in software engineering teaching to provide learners with reading materials of appropriate complexity. At present, the automatic classification of text complexity is still mainly based on the construction of linear model formulas. But due to the limited number of features eventually entering the model, the accuracy is generally not high, and it is difficult to extend to other data sets. This paper aims to explore the performance of neural network technology in the text complexity classification models with the help of multi-dimensional features and feature optimization. After comparative experiments, the text complexity classification model based on the neural network has the best performance. Its accuracy, recall, and F1 comprehensive evaluation indicators in cross-validation are better than other methods, which not only have higher prediction accuracy, but also have more stable performance. The model established from that has a strong generalization ability for new data, with obvious advantages.