Distributed Pattern Recognition Research Articles

Graph isomorphism network with weighted multi‐aggregators for building shape classification

AbstractBuilding shape cognition is essential for tasks, such as map generalization, urban modeling, and building semantics and distribution pattern recognition. Traditional geometric and statistical methods rely on human‐defined shape indicators, and spectral‐based graph neural networks (GNNs) require Laplacian eigendecomposition, resulting in high algorithmic complexity. Therefore, we proposed a low‐complexity and simple‐to‐use spatial‐domain GNN for differentiating building shapes. To examine the influence of the building vertices on their shape, we treated each building as a graph and proposed a graph isomorphic network with weighted multi‐aggregators (GIN‐WMA) by analyzing the node connectivity of a building graph. The GIN‐WMA utilizes a novel aggregator that combines the sum and max aggregators, enhancing its recognition and differentiation capabilities. This approach can effectively differentiate nodes that have identical features after aggregation by the sum aggregator. We extracted features considering both local node and global shape features, drawing inspiration from Gestalt cognitive psychology and GNN's “node–graph” differentiation strategy. In addition, we compared the performance of GIN‐WMA with existing methods, studying the effect of various node features and their combinations on classification accuracy. The results demonstrated that GIN‐WMA outperforms other methods in discriminating building shapes, demonstrating superior capabilities in shape classification and enabling end‐to‐end extraction and classification of building shapes.

Point Pattern Analysis (PPA) as a tool for reproducible archaeological site distribution analyses and location processes in early iron age south-west Germany.

Point Pattern Analysis (PPA) has gained momentum in archaeological research, particularly in site distribution pattern recognition compared to supra-regional environmental variables. While PPA is now a statistically well-established method, most of the data necessary for the analyses are not freely accessible, complicating reproducibility and transparency. In this article, we present a fully reproducible methodical framework to PPA using an open access database of archaeological sites located in south-west Germany and open source explanatory covariates to understand site location processes and patterning. The workflow and research question are tailored to a regional case study, but the code underlying the analysis is provided as an R Markdown file and can be adjusted and manipulated to fit any archaeological database across the globe. The Early Iron Age north of the Alps and particularly in south-west Germany is marked by numerous social and cultural changes that reflect the use and inhabitation of the landscape. In this work we show that the use of quantitative methods in the study of site distribution processes is essential for a more complete understanding of archaeological and environmental dynamics. Furthermore, the use of a completely transparent and easily adaptable approach can fuel the understanding of large-scale site location preferences and catchment compositions in archaeological, geographical and ecological research.

Construction of Art Resource Platform Based on Distributed Pattern Recognition SoC Deep Learning Algorithm

Construction of Art Resource Platform Based on Distributed Pattern Recognition SoC Deep Learning Algorithm

GraphLoc: a graph neural network model for predicting protein subcellular localization from immunohistochemistry images.

Recognition of protein subcellular distribution patterns and identification of location biomarker proteins in cancer tissues are important for understanding protein functions and related diseases. Immunohistochemical (IHC) images enable visualizing the distribution of proteins at the tissue level, providing an important resource for the protein localization studies. In the past decades, several image-based protein subcellular location prediction methods have been developed, but the prediction accuracies still have much space to improve due to the complexity of protein patterns resulting from multi-label proteins and the variation of location patterns across cell types or states. Here, we propose a multi-label multi-instance model based on deep graph convolutional neural networks, GraphLoc, to recognize protein subcellular location patterns. GraphLoc builds a graph of multiple IHC images for one protein, learns protein-level representations by graph convolutions and predicts multi-label information by a dynamic threshold method. Our results show that GraphLoc is a promising model for image-based protein subcellular location prediction with model interpretability. Furthermore, we apply GraphLoc to the identification of candidate location biomarkers and potential members for protein networks. A large portion of the predicted results have supporting evidence from the existing literatures and the new candidates also provide guidance for further experimental screening. The dataset and code are available at: www.csbio.sjtu.edu.cn/bioinf/GraphLoc. Supplementary data are available at Bioinformatics online.

End-to-End Calcification Distribution Pattern Recognition for Mammograms: An Interpretable Approach with GNN.

Purpose: We aimed to develop a novel interpretable artificial intelligence (AI) model algorithm focusing on automatic detection and classification of various patterns of calcification distribution in mammographic images using a unique graph convolution approach. Materials and methods: Images from 292 patients, which showed calcifications according to the mammographic reports and diagnosed breast cancers, were collected. The calcification distributions were classified as diffuse, segmental, regional, grouped, or linear. Excluded were mammograms with (1) breast cancer with multiple lexicons such as mass, asymmetry, or architectural distortion without calcifications; (2) hidden calcifications that were difficult to mark; or (3) incomplete medical records. Results: A graph-convolutional-network-based model was developed. A total of 581 mammographic images from 292 cases of breast cancer were divided based on the calcification distribution pattern: diffuse (n = 67), regional (n = 115), group (n = 337), linear (n = 8), or segmental (n = 54). The classification performances were measured using metrics including precision, recall, F1 score, accuracy, and multi-class area under the receiver operating characteristic curve. The proposed model achieved a precision of 0.522 ± 0.028, sensitivity of 0.643 ± 0.017, specificity of 0.847 ± 0.009, F1 score of 0.559 ± 0.018, accuracy of 64.325 ± 1.694%, and area under the curve of 0.745 ± 0.030; thus, the method was found to be superior compared to all baseline models. The predicted linear and diffuse classifications were highly similar to the ground truth, and the predicted grouped and regional classifications were also superior compared to baseline models. The prediction results are interpretable using visualization methods to highlight the important calcification nodes in graphs. Conclusions: The proposed deep neural network framework is an AI solution that automatically detects and classifies calcification distribution patterns on mammographic images highly suspected of showing breast cancers. Further study of the AI model in an actual clinical setting and additional data collection will improve its performance.

Mutational landscape of marginal zone B-cell lymphomas of various origin: organotypic alterations and diagnostic potential for assignment of organ origin

This meta-analysis aims to concisely summarize the genetic landscape of splenic, nodal and extranodal marginal zone lymphomas (MZL) in the dura mater, salivary glands, thyroid, ocular adnexa, lung, stomach and skin with respect to somatic variants. A systematic PubMed search for sequencing studies of MZL was executed. All somatic mutations of the organs mentioned above were combined, uniformly annotated, and a dataset containing 25 publications comprising 6016 variants from 1663 patients was created. In splenic MZL, KLF2 (18%, 103/567) and NOTCH2 (16%, 118/725) were the most frequently mutated genes. Pulmonary and nodal MZL displayed recurrent mutations in chromatin-modifier-encoding genes, especially KMT2D (25%, 13/51, and 20%, 20/98, respectively). In contrast, ocular adnexal, gastric, and dura mater MZL had mutations in genes encoding for NF-κB pathway compounds, in particular TNFAIP3, with 39% (113/293), 15% (8/55), and 45% (5/11), respectively. Cutaneous MZL frequently had FAS mutations (63%, 24/38), while MZL of the thyroid had a higher prevalence for TET2 variants (61%, 11/18). Finally, TBL1XR1 (24%, 14/58) was the most commonly mutated gene in MZL of the salivary glands. Mutations of distinct genes show origin-preferential distribution among nodal and splenic MZL as well as extranodal MZL at/from different anatomic locations. Recognition of such mutational distribution patterns may help assigning MZL origin in difficult cases and possibly pave the way for novel more tailored treatment concepts.

Recognition of Size Distribution Patterns in Fossil Coal

Recognition of Size Distribution Patterns in Fossil Coal

Recognition of Size Distribution Patterns in Fossil Coal

Recognition of Size Distribution Patterns in Fossil Coal

Application of graph theory and filter based variable selection methods in the design of a distributed data-driven monitoring system

Two methods that represent extensions of our previously developed methods for distributed data-driven monitoring are proposed. The first, Extended Forward Selection for Distributed Pattern Recognition, selects a decomposition for distributed pattern recognition such that diagnostic performance is near optimal subject to constraints. It uses a filter method to select sensors and allocates them among a minimum number of subsystems using graph theoretic algorithms. Its advantage over the Forward Selection for Distributed Pattern Recognition method is that it scales to systems with sensors in the order of 1,000. The second method, Extended Subsystem and Sensor Allocation, uses graph theoretic algorithms to find the minimum number of locations for distributed monitoring, the sensors that should transmit to each location, and the monitoring tasks at each location. Its main advantage over the original Subsystem and Sensor Allocation method is that it is applicable even when data is not available before plant operation begins.

Ficolin A derived from local macrophages and neutrophils protects against lipopolysaccharide-induced acute lung injury by activating complement.

Ficolins are important and widely distributed pattern recognition molecules that can induce lectin complement pathway activation and initiate the innate immune response. Although ficolins can bind lipopolysaccharide (LPS) in vitro, the sources, dynamic changes and roles of local ficolins in LPS-induced pulmonary inflammation and injury remain poorly understood. In this study, we established a ficolin knockout mouse model by clustered regularly interspaced short palindromic repeats/CRISPR-associated protein 9 (CRISPR/Cas9) technology, and used flow cytometry and hematoxylin and eosin staining to study the expressions and roles of local ficolins in LPS-induced pulmonary inflammation and injury. Our results show that besides ficolin B (FcnB), ficolin A (FcnA) is also expressed in leukocytes from the bone marrow, peripheral blood, lung and spleen. Further analyses showed that macrophages and neutrophils are the main sources of FcnA and FcnB, and T and B cells also express a small amount of FcnB. The intranasal administration of LPS induced local pulmonary inflammation with the increased recruitment of macrophages and neutrophils. LPS stimulation induced increased expression of FcnA and FcnB in neutrophils at the acute stage and in macrophages at the late stage. The severity of the lung injury and local inflammation of Fcna-/- mice was increased by the induction of extracellular complement activation. The recovery of LPS-induced local lung inflammation and injury was delayed in Fcnb-/- mice. Hence, these findings suggested that the local macrophage- and neutrophil-derived FcnA protects against LPS-induced acute lung injury by mediating extracellular complement activation.

Flow regime and volume fraction identification using nuclear techniques, artificial neural networks and computational fluid dynamics

Knowledge of the flow regime and the volume fraction in multiphase flow is of fundamental importance in predicting the performance of many systems and processes. This study is based on gamma-ray pulse height distribution pattern recognition by means of an artificial neural network. The detection system uses appropriate one narrow beam geometry, comprising a gamma-ray source and a NaI(Tl) detector. The models for annular and stratified flow regimes were developed using MCNPX code, in order to obtain adequate data sets for training and testing of the artificial neural network. Several experiments were carried out in the stratified flow regime to validate the simulated results. Finally, Ansys-CFX was used as computational fluid dynamics software to simulate two different volume fractions, which were modeled and transformed in voxels and transferred to MCNPX code. The use of computational fluid dynamics is of great importance, because it brings the studies closer to the reality. All flow regimes were correctly recognized and the volume fractions were appropriately predicted with relative errors less than 1.1%.

Data mining for recognition of spatial distribution patterns of building heights using airborne lidar data

There is an increasing demand for spatial big data visualisation in Geographic Information Systems (GIS) in building construction and urban development. Exploring building height patterns is required to obtain and visualize essential information about spatio-temporal vertical urban developments due to the trends towards increasing building heights in different urban fabrics. While metrics to characterize horizontal patterns of urban fabric using spectral information exist, theoritical-based metrics identifying the patterns of vertical urban developments using height values are still scarce. In addition, there is a lack of reliable methods to analyze height information for modeling the distribution of building heights and to automatically detect three-dimensional urban patterns. In this paper, we propose to apply the spatial statistics of Local Moran’s I (LMI), Gi∗ and Kernel Density Estimation (KDE) on building heights to explore vertical urban patterns through detecting the concentration of relatively higher buildings. The proposed methods were applied on two different airborne lidar point cloud data sets. The results show overall good performance of LMI and Gi* methods compared to KDE. It is also found that there is a higher level of agreement between clusters of relatively higher buildings derived by the autocorrelation statistics of LMI and Gi*, compared with the patterns derived from the Kernel density. For the lower accuracies obtained from the KDE, the authors suggest to use either LMI or Gi* for this kind of study. The spatial closeness of clusters of higher buildings to major roads, defined by the mean distances of the clusters to major roads, were investigated and based on the Analysis of Variance (ANOVA) and Tukey’s tests, the mean distances were found to be shorter than for all other buildings. Lastly, an analysis of clusters of the relatively higher buildings showed varying land uses for the two case studies.

Recognition of earthquake-induced landslide and spatial distribution patterns triggered by the Jiuzhaigou earthquake in August 8, 2017

2017年8月8日发生的7.0级九寨沟地震诱发九寨沟熊猫海附近产生大量的滑坡体，造成道路阻塞，严重影响地震应急救援进度。为快速准确地识别滑坡分布范围，本文在深入分析滑坡遥感影像特征的基础上，引入面向对象分析方法，实现了基于无人机影像的震后滑坡体的自动识别。通过多尺度分割算法获取滑坡多层次影像对象，利用SEaTH算法自动构建每一层次特征规则集，实现基于不同层次分析的滑坡体自动识别。分析滑坡体在地形、活动断层等因子中的空间分布特征，为地震滑坡预测与危险性评价奠定基础。与人工目视解译结果相比较，基于面向对象的滑坡自动识别方法提取精度可达94.8%，Kappa系数为0.827，在电脑配置相同的情况下，自动识别方法的效率是人工目视解译效率的一倍。空间分布特征分析表明，地震滑坡的空间分布与斜坡坡度、地形起伏度呈正相关关系，与地表粗糙度存在负相关关系，研究区滑坡体分布存在明显的断层效应。

Event Monitoring using Distributed Pattern Recognition Approach on Integrated IoT-Blockchain Network

With the advancement in the field of Internet-of-Things (IoT), event monitoring applications have rapidly evolved from a simple event data acquisition towards predictive event analytics involving multi-sensory data aggregation in a distributed environment. Existing event monitoring schemes are mainly relying on ine_cient centralized processing mechanism, which may lead to the common single-point of failure for the entire system. In addition, there is no proper method for verifying the event data generated by the monitoring system. In this paper, we present a distributed event monitoring scheme using a Hierarchical Graph Neuron (HGN) distributed pattern recognition algorithm. HGN is a single-cycle learning graph-based recognition scheme that is modelled for in-network deployment. In this work, event data retrieved from multi-sensory IoT devices within a distributed event monitoring network is converted into pattern. To address the event data verification problem, we integrate our proposed scheme with blockchain technology. Combining this IoT event monitoring capabilities with blockchain-based data storage and verification could leads towards a scalable event detection and monitoring model for large-scale network. The results obtained from our simulation shows that the proposed scheme o_ers high event detection accuracy and capable of minimizing the event storage requirements on blockchain network.

Density prediction for petroleum and derivatives by gamma-ray attenuation and artificial neural networks

This work presents a new methodology for density prediction of petroleum and derivatives for products' monitoring application. The approach is based on pulse height distribution pattern recognition by means of an artificial neural network (ANN). The detection system uses appropriate broad beam geometry, comprised of a 137Cs gamma-ray source and a NaI(Tl) detector diametrically positioned on the other side of the pipe in order measure the transmitted beam. Theoretical models for different materials have been developed using MCNP-X code, which was also used to provide training, test and validation data for the ANN. 88 simulations have been carried out, with density ranging from 0.55 to 1.26gcm−3 in order to cover the most practical situations. Validation tests have included different patterns from those used in the ANN training phase. The results show that the proposed approach may be successfully applied for prediction of density for these types of materials. The density can be automatically predicted without a prior knowledge of the actual material composition.

Patterns of 11C-PIB cerebral retention in mild cognitive impairment patients

ObjectiveTo evaluate the patterns of cerebral cortical distribution of 11C-PIB in patients with mild cognitive impairment (MCI). Material and methodsThe study included 69 patients (37 male, age range 42–79 years) with MCI, sub-classified as 53 with amnestic-MCI (A-MCI), and 16 with non-amnestic-MCI (NA-MCI). Patients underwent 11C-PIB PET/CT scan 60min after intravenous injection of the radiotracer. A visual analysis of the images was performed by 2 experienced physicians. 11C-PIB-positive studies were considered when gray matter uptake was equal to or greater than white matter. According to the regions involved, 11C-PIB-positive studies were classified into A-pattern (predominant retention in frontal, anterior cingulate, lateral temporal, and basal ganglia) and B-pattern (generalized retention). ResultsThirty-nine of the 69 (56%) patients with MCI showed 11C-PIB retention. Of the 53 A-MCI patients, 36 (68%) showed 11C-PIB retention. Eleven out of 36 (30%) positive scans in A-MCI patients showed A-pattern, and 25 out of 36 (70%) patients had a B-pattern. Positive 11C-PIB was observed in 3 out of 16 (19%) patients with NA-MCI. Regional distribution in these 3 patients showed A-pattern in 1, and B-pattern in 2 patients. ConclusionCortical retention of 11C-PIB was more frequent in A-MCI than in NA-MCI patients, and also B-pattern than A-pattern in the 11C-PIB positive group. The recognition of 11C-PIB distribution patterns allows MCI patients to be classified, and the A-pattern may offer a therapeutic window for potential future treatments.

Research on Normal Human Plantar Pressure Test

FSR400 pressure sensor, nRF905 wireless transceiver and MSP40 SCM are used to design the insole pressure collection system, LabVIEW is used to make HMI of data acquisition, collecting a certain amount of normal human foot pressure data, statistical analysis of pressure distribution relations about five stages of swing phase during walking, using the grid closeness degree to identify plantar pressure distribution pattern recognition, and the algorithm simulation, experimental results demonstrated this method feasible.

Patterns of 11C-PIB cerebral retention in mild cognitive impairment patients.

To evaluate the patterns of cerebral cortical distribution of (11)C-PIB in patients with mild cognitive impairment (MCI). The study included 69 patients (37 male, age range 42-79 years) with MCI, sub-classified as 53 with amnestic-MCI (A-MCI), and 16 with non-amnestic-MCI (NA-MCI). Patients underwent (11)C-PIB PET/CT scan 60min after intravenous injection of the radiotracer. A visual analysis of the images was performed by 2 experienced physicians. (11)C-PIB-positive studies were considered when gray matter uptake was equal to or greater than white matter. According to the regions involved, (11)C-PIB-positive studies were classified into A-pattern (predominant retention in frontal, anterior cingulate, lateral temporal, and basal ganglia) and B-pattern (generalized retention). Thirty-nine of the 69 (56%) patients with MCI showed (11)C-PIB retention. Of the 53 A-MCI patients, 36 (68%) showed (11)C-PIB retention. Eleven out of 36 (30%) positive scans in A-MCI patients showed A-pattern, and 25 out of 36 (70%) patients had a B-pattern. Positive (11)C-PIB was observed in 3 out of 16 (19%) patients with NA-MCI. Regional distribution in these 3 patients showed A-pattern in 1, and B-pattern in 2 patients. Cortical retention of (11)C-PIB was more frequent in A-MCI than in NA-MCI patients, and also B-pattern than A-pattern in the (11)C-PIB positive group. The recognition of (11)C-PIB distribution patterns allows MCI patients to be classified, and the A-pattern may offer a therapeutic window for potential future treatments.

Information flow in layered networks of non-monotonic units

Layered neural networks are feedforward structures that yield robust parallel and distributed pattern recognition. Even though much attention has been paid to pattern retrieval properties in such systems, many aspects of their dynamics are not yet well characterized or understood. In this work we study, at different temperatures, the memory activity and information flows through layered networks in which the elements are the simplest binary odd non-monotonic function. Our results show that, considering a standard Hebbian learning approach, the network information content has its maximum always at the monotonic limit, even though the maximum memory capacity can be found at non-monotonic values for small enough temperatures. Furthermore, we show that such systems exhibit rich macroscopic dynamics, including not only fixed point solutions of its iterative map, but also cyclic and chaotic attractors that also carry information.

GIS technology in regional recognition of the distribution pattern of multifloral honey: The chemical traits in Serbia

GIS is a computer-based system to input, store, manipulate, analyze and output spatially referenced data. There is a huge range application of GIS that generally sets out to fulfill: mapping, measurement, monitoring, modeling and management. In this study, GIS technology was used for the regional recognition of origin and distribution patterns of multifloral honey chemical traits in Serbia. This included organizing and analyzing the spatial and attributive data of 164 honey samples collected from different regions of Serbia during the harvesting season of 2009. Multifloral honey was characterized in regards to mineral composition, sugar content and basic physicochemical properties. The kriging method of Geostatistical Analyst was used for interpolation to predict values of a sampled variable over the whole territory of Serbia.