This article proposes a novel community detection approach, MHNMF, which analyzes the multihop connectivity patterns within the network. Subsequently, we devise an optimized algorithm to enhance MHNMF, coupled with a theoretical investigation into its computational intricacy and convergence patterns. The performance of MHNMF on 12 actual benchmark networks was assessed against 12 existing community detection methods, demonstrating that MHNMF is superior in performance.
Inspired by the human visual system's global-local processing, we propose a novel convolutional neural network (CNN), CogNet, which comprises a global pathway, a local pathway, and a top-down modulation mechanism. A common CNN block is first applied to establish the local pathway, which has the task of extracting detailed local features from the input image. Following this, we leverage a transformer encoder to construct the global pathway, enabling us to capture the global structural and contextual information inherent in the local parts of the input image. The culminating stage entails the construction of a learnable top-down modulator that fine-tunes the local features of the local pathway using global information from the global pathway. To simplify usage, we encapsulate the dual-pathway computation and modulation procedure into a fundamental component, the global-local block (GL block). A CogNet of any depth can be built by concatenating a requisite number of GL blocks. Evaluations of the proposed CogNets on six benchmark datasets consistently achieved leading-edge accuracy, showcasing their effectiveness in overcoming texture bias and resolving semantic confusion encountered by traditional CNN models.
During the process of walking, human joint torques are commonly determined through the application of inverse dynamics. Ground reaction force and kinematic measurements are prerequisites for analysis in traditional approaches. A novel real-time hybrid approach is introduced herein, merging a neural network and a dynamic model, requiring only kinematic data for operation. For direct joint torque estimation, a neural network model spanning the input of kinematic data to the output is created. A diverse set of walking conditions, including the initiation and cessation of movement, unexpected alterations in speed, and one-sided gaits, fuel the training of the neural networks. The first test of the hybrid model involved a detailed dynamic gait simulation in OpenSim, ultimately achieving root mean square errors under 5 N.m and a correlation coefficient over 0.95 for all the joints. Empirical studies show that the end-to-end model typically performs better than its hybrid counterpart across the complete testing regime, in comparison with the benchmark established by the gold standard, which incorporates both kinetic and kinematic aspects. The two torque estimators were likewise evaluated in a single participant, while wearing a lower limb exoskeleton. This instance showcases the hybrid model (R>084) performing considerably better than the end-to-end neural network (R>059). Sodium dichloroacetate chemical structure Scenarios that diverge from the training data are more effectively addressed by the superior hybrid model.
Thromboembolism's progression within blood vessels, if left uncontrolled, may cause life-threatening conditions such as stroke, heart attack, and even sudden death. Thromboembolism treatment, with sonothrombolysis augmented by ultrasound contrast agents, displays encouraging outcomes. A novel treatment for deep vein thrombosis, intravascular sonothrombolysis, has recently been highlighted for its potential to be both effective and safe. Despite the encouraging results from the treatment, optimal clinical application efficiency may not be achieved due to the lack of imaging guidance and clot characterization in the thrombolysis procedure. A miniaturized intravascular sonothrombolysis transducer, constructed from an 8-layer PZT-5A stack having a 14×14 mm² aperture, was designed and assembled into a custom two-lumen 10-Fr catheter, as detailed in this paper. The treatment's progress was tracked using internal-illumination photoacoustic tomography (II-PAT), a hybrid imaging method that merges optical absorption's robust contrast with ultrasound's deep detection capabilities. Through intravascular light delivery facilitated by a thin optical fiber integrated with the catheter, II-PAT effectively overcomes the optical attenuation-induced limitations on tissue penetration depth. PAT-guided in-vitro sonothrombolysis experiments involved synthetic blood clots, which were placed within a tissue phantom. Using a clinically significant depth of ten centimeters, the II-PAT system can estimate the oxygenation level, position, stiffness, and shape of clots. Polyhydroxybutyrate biopolymer Our findings reveal the feasibility of the proposed PAT-guided intravascular sonothrombolysis, with a real-time feedback mechanism actively implemented during the treatment.
This study presents a computer-aided diagnosis (CADx) framework, CADxDE, designed for dual-energy spectral CT (DECT) applications. CADxDE operates directly on the transmission data in the pre-log domain to analyze spectral information for lesion identification. The CADxDE is equipped with material identification and machine learning (ML)-powered CADx functionality. DECT's virtual monoenergetic imaging of identified materials allows machine learning to study the responses of different tissue types (such as muscle, water, and fat) within lesions at each corresponding energy level, ultimately aiding computer-aided diagnosis (CADx). Preserving the essential information in the DECT scan, an iterative reconstruction process using a pre-log domain model is applied to generate decomposed material images. These images subsequently produce virtual monoenergetic images (VMIs) at predetermined n energies. Although these VMIs share the same anatomical structure, their contrasting distributions reveal intricate details, providing valuable information for tissue characterization, along with the associated n-energies. Therefore, a corresponding machine learning-driven CADx system is developed to capitalize on the energy-amplified tissue attributes for the discrimination of malignant and benign lesions. Oral medicine An innovative multi-channel 3D convolutional neural network (CNN) approach, operating on original images and utilizing machine learning (ML) methods based on extracted lesion features, is designed to showcase the viability of CADxDE. Three pathologically verified clinical data sets demonstrated a substantial improvement in AUC scores, ranging from 401% to 1425% higher than conventional DECT data (high and low energy) and conventional CT data. CADxDE's innovative energy spectral-enhanced tissue features contributed to a marked enhancement of lesion diagnosis performance, as indicated by a mean AUC gain greater than 913%.
The task of classifying whole-slide images (WSI) in computational pathology is crucial, but faces substantial obstacles including the extremely high resolution, the high cost of manual annotation, and data heterogeneity. Despite its potential in whole-slide image (WSI) classification, multiple instance learning (MIL) struggles with memory limitations imposed by the gigapixel resolution. A common approach in existing MIL networks to address this issue is to isolate the feature encoder from the MIL aggregator, although this separation may lead to significant drops in performance. This paper's Bayesian Collaborative Learning (BCL) framework aims to resolve the memory bottleneck challenge presented by WSI classification. To achieve collaborative learning of the feature encoder and MIL aggregator within the target MIL classifier, we introduce an auxiliary patch classifier. This interaction will prevent the memory bottleneck. In a unified Bayesian probabilistic framework, a collaborative learning procedure is developed, and a principled Expectation-Maximization algorithm is applied to infer the optimal model parameters iteratively. For an effective implementation of the E-step, a pseudo-labeling method that considers quality is also presented. Using CAMELYON16, TCGA-NSCLC, and TCGA-RCC datasets, the proposed BCL was evaluated, achieving AUC scores of 956%, 960%, and 975% respectively. This performance consistently surpasses all other comparative methods. A presentation of the method's in-depth analysis and discussion will be provided to enhance comprehension. For prospective work, we have made our source code accessible at https://github.com/Zero-We/BCL.
Anatomical representation of head and neck vessels serves as a pivotal diagnostic step in cerebrovascular disease evaluation. Precise and automated vessel labeling in computed tomography angiography (CTA) continues to be a complex task, especially for the head and neck vasculature, where vessels are tortuous, branched, and frequently situated close to other vasculature. To combat these difficulties, we introduce a novel topology-cognizant graph network, TaG-Net, for the application of vessel labeling. This approach combines the strengths of volumetric image segmentation in the voxel space and centerline labeling in the line space, ensuring detailed local features from the voxel space and superior anatomical and topological vessel data from the vascular graph created from centerlines. Centerlines from the initial vessel segmentation are extracted, and a vascular graph is then constructed. Finally, vascular graph labeling is performed using TaG-Net, which consists of topology-preserving sampling, topology-aware feature grouping, and multi-scale vascular graph approaches. Following this, the vascular graph, marked with labels, is used to enhance volumetric segmentation by completing vessel structures. The final step involves labeling the head and neck vessels of 18 segments, achieved by applying centerline labels to the refined segmentation. Comparative analysis of CTA images from 401 subjects underscores our method's superior vessel segmentation and labeling, showcasing an advancement over current state-of-the-art techniques.
Real-time inference is a key benefit of regression-based multi-person pose estimation, which is gaining significant traction.