Related Features Research Articles

Lifelong psychological and lifestyle features in relation to cardiovascular disease development: A narrative review

Despite the extensive literature on individual risk factors, such as diet, physical activity, body weight, smoking habits, and mental health characteristics, the literature on different lifestyle and mental health patterns, their evolution over time, and how these factors interact and synergistically affect cardiovascular health is sparse. The need for longitudinally studying these characteristics is currently being highlighted by the limited studies with repeated measurements. This review summarizes the literature concerning lifestyle and psychological characteristics, both well-studied (e.g., diet, physical activity, smoking habits, chronic stress and depression), as well as newly proposed characteristics (e.g., sleep habits, positive psychology characteristics) and their longitudinal association with cardiovascular disease. Studies with data on lifelong characteristics are not always comparable and have found mixed results, especially concerning changes in characteristics such as quitting smoking or starting to be physically active later in life. Therefore, further research is warranted for the implementation of future preventive strategies for cardiovascular disease.

Multi-feature Learning Adaptive Network for Underwater Image Enhancement

Underwater image enhancement faces variety of challenges owing to the diversity of underwater scenes (viewed as water types) and the rich multi-frequency information. To deal with these challenges, this paper proposes a multi-feature learning adaptive underwater image enhancement network comprising an adaptive module and a dual-layer synchronous enhancement network. First, we design an adaptive module which enables the determination of water type inside the model and eliminates the negative effect of water type diversity by building water type related features. Then, the model learns high-frequency and low-frequency features through a dual-layer synchronous enhancement network to extract more comprehensive information. Finally, the outputs of the dual-layer network are merged to obtain more realistic underwater enhanced images. Numerous experiments have shown that the proposed method outperforms the comparison method for visual perception and assessment metrics.

Identifying suicide attempter in major depressive disorder through machine learning: the importance of pain avoidance, event-related potential features of affective processing.

How to achieve a high-precision suicide attempt classifier based on the three-dimensional psychological pain model is a valuable issue in suicide research. The aim of the present study is to explore the importance of pain avoidance and its related neural features in suicide attempt classification models among patients with major depressive disorder. By recursive feature elimination with cross-validation and support-vector-machine algorithms, scores from the measurements and the task-based EEG signals were chosen to achieve a suicide attempt classification model. In the multimodal suicide attempt classifier with an accuracy of 83.91% and an area under the curve of 0.90, pain avoidance ranked as the top one in the optimal feature set. Theta (reward positive feedback minus neutral positive feedback) was the shared neural representation ranking as the top one of event-related potential features in pain avoidance and suicide attempt classifiers. In conclusion, the suicide attempt classifier based on pain avoidance and its related affective processing neural features has excellent accuracy among patients with major depressive disorder. Pain avoidance is a stable and strong indicator for identifying suicide risks in both traditional analyses and machine-learning approaches. A novel methodology is needed to clarify the relationship between cognitive and affective processing evoked by punishment stimuli and pain avoidance.

The Neighbourhood–Health Nexus: Design, Behaviour and Futures

Over the last decade, and more recently with the coronavirus disease 2019 (COVID- 19) pandemic, increased attention has been given to the dynamic between health and urbanism. Features such as city design, the environment, and socioeconomic factors have been studied worldwide. Most studies have focused on a single element of the urban environment, making it difficult to understand the possible influence of related urban features. Furthermore, studies have addressed the issue of urbanism and health on different international, national, urban, and local scales, resulting in multiple inconsistencies. With the enhanced growth of cities, it is argued that the neighbourhood scale is the ideal scale to understand the built environmental–health nexus. More specifically, the paper reviews studies that focus on neighbourhood design and its influence on health, and studies that focus on residents' health-related behaviour. In addition, it maps the key developments in e-health and its expected influence on health services in neighbourhoods. Insights from these reviews are used to offer a preliminary conceptual framework for addressing health in neighbourhoods.

Fast Multilabel Feature Selection via Global Relevance and Redundancy Optimization.

Information theoretical-based methods have attracted a great attention in recent years and gained promising results for multilabel feature selection (MLFS). Nevertheless, most of the existing methods consider a heuristic way to the grid search of important features, and they may also suffer from the issue of fully utilizing labeling information. Thus, they are probable to deliver a suboptimal result with heavy computational burden. In this article, we propose a general optimization framework global relevance and redundancy optimization (GRRO) to solve the learning problem. The main technical contribution in GRRO is a formulation for MLFS while feature relevance, label relevance (i.e., label correlation), and feature redundancy are taken into account, which can avoid repetitive entropy calculations to obtain a global optimal solution efficiently. To further improve the efficiency, we extend GRRO to filter out inessential labels and features, thus facilitating fast MLFS. We call the extension as GRROfast, in which the key insights are twofold: 1) promising labels and related relevant features are investigated to reduce ineffective calculations in terms of features, even labels and 2) the framework of GRRO is reconstructed to generate the optimal result with an ensemble. Moreover, our proposed algorithms have an excellent mechanism for exploiting the inherent properties of multilabel data; specifically, we provide a formulation to enhance the proposal with label-specific features. Extensive experiments clearly reveal the effectiveness and efficiency of our proposed algorithms.

Classification of the Cardiac Arrhythmia Using Combined Feature Selection Algorithms

The prediction of heart disease has gained great importance in recent years. Efficient monitoring of cardiac patients can save tremendous number of lives. This paper presents a method for classification and prediction of electrocardiogram data obtained from 452 patients representing the risk of cardiac arrhythmia. The aim of the study is to select highly related features with arrhythmia risk by using three different feature selection algorithms. In addition, various machine learning models are utilized for the classification task such as k-Nearest Neighbors (k-NN), Support Vector Machines (SVM) and Decision Tree (DT). The experimental results show that combination of a purposed feature selection method which later is called “Matched Selection” using SVM classifier outperforms other combinations and have an accuracy of 81.27% while k-NN and DT classifiers have an accuracy of 69.66% and 73.50% respectively. The study, in which detailed analyses are presented comparatively, is promising for the future studies.

An MRI radiomics approach to discriminate hemorrhage prone intracranial tumors before stereotactic biopsy.

To explore imaging biomarkers predictive of intratumoral hemorrhage for lesions intended for elective stereotactic biopsy. This study included a retrospective cohort of 143 patients with 175 intracranial lesions intended for stereotactic biopsy. All the lesions were randomly split into a training dataset (n=121) and a test dataset (n=54) at a ratio of 7:3. 34 lesions were defined as "hemorrhage-prone tumors" as hemorrhage occurred between initial diagnostic MRI acquisition and the scheduled biopsy procedure. Radiomics features were extracted from the contrast-enhanced T1WI and T2WI images. Features informative of hemorrhage were then selected by the LASSO algorithm and an SVM model was built with selected features. The SVM model was further simplified by discarding features with low importance calculated using a "permutation importance" method. The model's performance was evaluated with confusion matrix-derived metrics and AUC value on the independent test dataset. Nine radiomics features were selected as hemorrhage related features of intracranial tumors by the LASSO algorithm. The simplified model's sensitivity, specificity, accuracy, and AUC reached 0.909, 0.930, 0.926, and 0.949 (95%CI: 0.865-1.000) on the test dataset in the discrimination of "hemorrhage-prone tumors". The permutation method rated feature "T2_gradient_firstorder_10Percentile" as the most important, the absence of which decreased the model's accuracy by 10.9%. Radiomics features extracted on contrast-enhanced T1WI and T2WI sequences were predictive of future hemorrhage of intracranial tumors with favorable accuracy. This model may assist in the arrangement of biopsy procedures and the selection of target lesions in patients with multiple lesions.

A Novel Missense Variant in the CHST3 Underlies Spondyloepiphyseal Dysplasia with Congenital Joint Dislocations

Background: Spondyloepiphyseal dysplasia (SED) is characterized by skeletal dysplasia and multiple joint dislocations. SEDs encompass various types, such as SED congenita, SED tarda (SED-T), SED with congenital joint dislocations (SED-CJD), SED stanescu, and SED-T with progressive arthropathy. Methods and Results: In the present study, we clinically and genetically characterized a consanguineous Pakistani family with SED-CJD. The affected member showed large joint dislocation, spinal deformities, and previously unreported facial features. Exome sequencing followed by Sanger sequencing revealed a missense variant, [c.601T>A; p.(Tyr201Asn)], in the CHST3. Conclusion: This study has not only expended the mutation spectrum in the gene CHST3 but also will facilitate diagnosis and genetic counseling of related features in the Pakistani population.

ScMNMF: a novel method for single-cell multi-omics clustering based on matrix factorization.

The technology for analyzing single-cell multi-omics data has advanced rapidly and has provided comprehensive and accurate cellular information by exploring cell heterogeneity in genomics, transcriptomics, epigenomics, metabolomics and proteomics data. However, because of the high-dimensional and sparse characteristics of single-cell multi-omics data, as well as the limitations of various analysis algorithms, the clustering performance is generally poor. Matrix factorization is an unsupervised, dimensionality reduction-based method that can cluster individuals and discover related omics variables from different blocks. Here, we present a novel algorithm that performs joint dimensionality reduction learning and cell clustering analysis on single-cell multi-omics data using non-negative matrix factorization that we named scMNMF. We formulate the objective function of joint learning as a constrained optimization problem and derive the corresponding iterative formulas through alternating iterative algorithms. The major advantage of the scMNMF algorithm remains its capability to explore hidden related features among omics data. Additionally, the feature selection for dimensionality reduction and cell clustering mutually influence each other iteratively, leading to a more effective discovery of cell types. We validated the performance of the scMNMF algorithm using two simulated and five real datasets. The results show that scMNMF outperformed seven other state-of-the-art algorithms in various measurements. scMNMF code can be found at https://github.com/yushanqiu/scMNMF.

Histopathological comparison of Sjögren-related features between paired labial and parotid salivary gland biopsies of sicca patients.

To compare focus score and other histopathological features between paired labial and parotid salivary gland biopsies in a diagnostic cohort of suspected Sjögren's disease (SjD) patients. Labial and parotid salivary gland biopsies were simultaneously obtained from patients with sicca complaints, suspected of having SjD. Biopsies were formalin fixed and paraffin embedded. Sections were stained with haematoxylin & eosin, and for CD3, CD20, CD45, cytokeratin, CD21, Bcl6, activation-induced deaminase (AID) and IgA/IgG. Focus score and other histopathological features characteristic for SjD were analysed. Based on the expert opinion of three experienced rheumatologists, 36 patients were diagnosed as SjD and 63 as non-SjD sicca patients. When taking all patients together, absolute agreement of various histopathological features between labial and parotid biopsies was high and varied between 80% (focus score) and 93% [(pre-)lymphoepithelial lesions (LELs)]. More labial gland biopsies had a focus score ≥1 compared with their parotid counterpart. Accordingly, the area of infiltrate was larger in labial gland biopsies. When considering only SjD patients, labial glands contained significantly fewer B-lymphocytes and germinal centres/mm2, and less severe LELs compared with parotid glands. Labial and parotid glands from SjD patients contain similar histopathological key features, and thus both glands can be used for diagnosis and classification of SjD. However, parotid salivary glands reveal more evident B-lymphocyte-related features, while labial glands exhibit more inflammation, which may be partially unrelated to SjD.

Boosting Few-Shot Learning via Attentive Feature Regularization

Few-shot learning (FSL) based on manifold regularization aims to improve the recognition capacity of novel objects with limited training samples by mixing two samples from different categories with a blending factor. However, this mixing operation weakens the feature representation due to the linear interpolation and the overlooking of the importance of specific channels. To solve these issues, this paper proposes attentive feature regularization (AFR) which aims to improve the feature representativeness and discriminability. In our approach, we first calculate the relations between different categories of semantic labels to pick out the related features used for regularization. Then, we design two attention-based calculations at both the instance and channel levels. These calculations enable the regularization procedure to focus on two crucial aspects: the feature complementarity through adaptive interpolation in related categories and the emphasis on specific feature channels. Finally, we combine these regularization strategies to significantly improve the classifier performance. Empirical studies on several popular FSL benchmarks demonstrate the effectiveness of AFR, which improves the recognition accuracy of novel categories without the need to retrain any feature extractor, especially in the 1-shot setting. Furthermore, the proposed AFR can seamlessly integrate into other FSL methods to improve classification performance.

Applying image features of proximal paracancerous tissues in predicting prognosis of patients with hepatocellular carcinoma

BackgroundMost of the methods using digital pathological image for predicting Hepatocellular carcinoma (HCC) prognosis have not considered paracancerous tissue microenvironment (PTME), which are potentially important for tumour initiation and metastasis. This study aimed to identify roles of image features of PTME in predicting prognosis and tumour recurrence of HCC patients. MethodsWe collected whole slide images (WSIs) of 146 HCC patients from Sun Yat-sen Memorial Hospital (SYSM dataset). For each WSI, five types of regions of interests (ROIs) in PTME and tumours were manually annotated. These ROIs were used to construct a Lasso Cox survival model for predicting the prognosis of HCC patients. To make the model broadly useful, we established a deep learning method to automatically segment WSIs, and further used it to construct a prognosis prediction model. This model was tested by the samples of 225 HCC patients from the Cancer Genome Atlas Liver Hepatocellular Carcinoma (TCGA-LIHC). ResultsIn predicting prognosis of the HCC patients, using the image features of manually annotated ROIs in PTME achieved C-index 0.668 in the SYSM testing dataset, which is higher than the C-index 0.648 reached by the model only using image features of tumours. Integrating ROIs of PTME and tumours achieved C-index 0.693 in the SYSM testing dataset. The model using automatically segmented ROIs of PTME and tumours achieved C-index of 0.665 (95% CI: 0.556–0.774) in the TCGA-LIHC samples, which is better than the widely used methods, WSISA (0.567), DeepGraphSurv (0.593), and SeTranSurv (0.642). Finally, we found the Texture SumAverage Skew HV on immune cell infiltration and Texture related features on desmoplastic reaction are the most important features of PTME in predicting HCC prognosis. We additionally used the model in prediction HCC recurrence for patients from SYSM-training, SYSM-testing, and TCGA-LIHC datasets, indicating the important roles of PTME in the prediction. ConclusionsOur results indicate image features of PTME is critical for improving the prognosis prediction of HCC. Moreover, the image features related with immune cell infiltration and desmoplastic reaction of PTME are the most important factors associated with prognosis of HCC.

A data-driven newsvendor model for elective-emergency admission control under uncertain inpatient bed capacity.

Elective-emergency admission control referred to allocating available inpatient bed capacity between elective and emergency hospitalization demand. Existing approaches for admission control often excluded several complex factors when making decisions, such as uncertain bed capacity and unknown true probability distributions of patient arrivals and departures. We aimed to create a data-driven newsvendor framework to study the elective-emergency admission control problem to achieve bed operational efficiency and effectiveness. We developed a data-driven approach that utilized the newsvendor framework to formulate the admission control problem. We also created approximation algorithms to generate a pool of candidate admission control solutions. Past observations and relevant emergency demand and bed capacity features were modeled in a newsvendor framework. Using approximation algorithmic approaches (sample average approximation, separated estimation and optimization, linear programing-LP, and distribution-free model) allowed us to derive computationally efficient data-driven solutions with tight bounds on the expected in-sample and out-of-sample cost guaranteed. Tight generalization bounds on the expected out-of-sample cost of the feature-based model were derived with respect to the LP and quadratic programing (QP) algorithms, respectively. Results showed that the optimal feature-based model outperformed the optimal observation-based model with respect to the expected cost. In a setting where the unit overscheduled cost was higher than the unit under-scheduled cost, scheduling fewer elective patients would replace the benefit of incorporating related features in the model. The tighter the available bed capacity for elective patients, the bigger the difference of the schedule cost between the feature-based model and the observation-based model. The study provides a reference for the theoretical study on bed capacity allocation between elective and emergency patients under the condition of the unknown true probability distribution of bed capacity and emergency demand, and it also proves that the approximate optimal policy has good performance.

Russian Practice of Applying Hedge Accounting

Currently there are no comprehensive empirical studies in the scientific literature regarding hedge accounting in non-financial organizations of the Russian Federation for the period from 2010 to 2021, which does not allow us to identify trends, problematic aspects and potential directions for the development of accounting technology in close connection with real practice. The purpose of the study is to analyze the results of scientific work carried out on this topic. The author of the article substantiates the need to analyze Russian practice in the field of hedge accounting, and also studies the experience of its application, including analysis in the context of separate companies. The theoretical basis of the study was the provisions of IFRS and US GAAP in terms of derivatives and hedge accounting, literature sources on the topic and the empirical basis — is the annual consolidated financial statements of 22 Russian companies for the period from 2010 to 2021. It has been revealed that the main thing in non-financial organizations is cash flow hedge accounting. It has been established that there is a gap between the facts of risk reduction or mitigation based on derivatives and the application of hedge accounting to them, which is evidence of the conscious refusal of some companies to apply this accounting methodology. The analysis of the identified cases of hedge accounting shows the examples of using exchange differences for hedging (rather than derivatives); also, the adaptation of fair value hedging to the practice of non-financial organizations is considered. The results of the study will be useful for non-financial organizations that have not applied, but consider hedging and related accounting features as a promising area of activity; researchers and consultants in the field of futures and derivatives market development and risk management.

A Retrospective Analysis of the Computed Tomography Findings and Diagnosis of 53 Cases of Elastofibroma in the Infrascapular Region.

In this work, we have used histopathology as the gold standard for the diagnosis, calculated the sensitivity and positive predictive value (PPV) of computed tomography (CT), and analyzed the CT and clinical characteristics of pathologically proven elastofibromas. A systematic retrospective analysis was performed on all patients with infrascapular lesions who were treated in the hospital from 2006 to 2018. CT and histopathological examinations were performed for all cases, and the CT sensitivity and PPV for the diagnosis of elastofibroma were calculated. 12 of 53 cases (20 lesions) underwent enhanced CT scan after CT plain scan, and the related clinical and CT features of elastofibromas have been discussed. Of the 54 patients treated during the study, CT diagnosis was consistent with histopathology in 53 cases. One was a false-positive patient. The PPV and sensitivity of the CT in the diagnosis of elastofibroma were 93.3% (95% CI 68.0%-99.8%) and 100%, respectively. The CT values of 12 patients with 20 lesions on plain and enhanced scans were statistically significant (P=0.001). The prevalence of elastofibromas in males and females was statistically significant (P=.000). There was no statistically significant difference in the incidence of left and right elastofibromas (P=0.752). There was no significant difference in the volume of left and right lesions (P=0.209) and the volume of elastofibromas between males and females (P=.474). CT is the most practical tool for the evaluation of elastofibromas in the infrascapular region.

Diagnostic yield and clinical impact of germline sequencing in children with CNS and extracranial solid tumors—a nationwide, prospective Swedish study

Childhood cancer predisposition (ChiCaP) syndromes are increasingly recognized as contributing factors to childhood cancer development. Yet, due to variable availability of germline testing, many children with ChiCaP might go undetected today. We report results from the nationwide and prospective ChiCaP study that investigated diagnostic yield and clinical impact of integrating germline whole-genome sequencing (gWGS) with tumor sequencing and systematic phenotyping in children with solid tumors. gWGS was performed in 309 children at diagnosis of CNS (n=123, 40%) or extracranial (n=186, 60%) solid tumors and analyzed for disease-causing variants in 189 known cancer predisposing genes. Tumor sequencing data were available for 74% (227/309) of patients. In addition, a standardized clinical assessment for underlying predisposition was performed in 95% (293/309) of patients. The prevalence of ChiCaP diagnoses was 11% (35/309), of which 69% (24/35) were unknown at inclusion (diagnostic yield 8%, 24/298). A second-hit and/or relevant mutational signature was observed in 19/21 (90%) tumors with informative data. ChiCaP diagnoses were more prevalent among patients with retinoblastomas (50%, 6/12) and high-grade astrocytomas (37%, 6/16), and in those with non-cancer related features (23%, 20/88), and ≥2 positive ChiCaP criteria (28%, 22/79). ChiCaP diagnoses were autosomal dominant in 80% (28/35) of patients, yet confirmed de novo in 64% (18/28). The 35 ChiCaP findings resulted in tailored surveillance (86%, 30/35) and treatment recommendations (31%, 11/35). Overall, our results demonstrate that systematic phenotyping, combined with genomics-based diagnostics of ChiCaP in children with solid tumors is feasible in large-scale clinical practice and critically guides personalized care in a sizable proportion of patients. The study was supported by the Swedish Childhood Cancer Fund and the Ministry of Health and Social Affairs.

Occupational Structure of Population in the Malaprabha River Basin, Karnataka State, India; A Geographical Approach

Changes in the size, composition, and distribution of the population are closely associated with the demographic structure of the workforce. On the other hand, the workforce participation rates vary according to the stages of economic development, across cultures, age groups, and between sexes (R. B. Bhagat and K. C. Das, 2008). It is an indicator of a growing society (Pant, 1992). Thus, occupational structure is the most important demographic aspect in explaining the economic well-being structure of the inhabitantsof the region. The occupational structure, in turn, is influenced by the participation rates and related features, such as the growth and levels of farm and rural non-farm activities. An attempt is made in this study to examine the emerging trends in the occupational structure of the population in the Malaprabha River Basin area from 1971 to 2011. More specifically, the spatio-temporal analysis of the work participation rates in total and farm sectors among males and females in the talukas of the study area This study is based upon secondary sources of data, and though the study area is a natural region, the talukas or tehsils have been taken as units of study. The findings suggest that there are undoubtedly significant changes in work participation rates between farm and non-farm sectors and between males and females in the study area. To overcome this context, some developmental policies such as youth employment is one of the prime focus areas of SHGs. This can bring prosperity to workers in far-off rural areas.

AdverSPAM: Adversarial SPam Account Manipulation in Online Social Networks

In recent years, the widespread adoption of Machine Learning (ML) at the core of complex IT systems has driven researchers to investigate the security and reliability of ML techniques. A very specific kind of threats concerns the adversary mechanisms through which an attacker could induce a classification algorithm to provide the desired output. Such strategies, known as Adversarial Machine Learning (AML), have a twofold purpose: to calculate a perturbation to be applied to the classifier’s input such that the outcome is subverted, while maintaining the underlying intent of the original data. Although any manipulation that accomplishes these goals is theoretically acceptable, in real scenarios perturbations must correspond to a set of permissible manipulations of the input, which is rarely considered in the literature. In this article, we present AdverSPAM , an AML technique designed to fool the spam account detection system of an Online Social Network (OSN). The proposed black-box evasion attack is formulated as an optimization problem that computes the adversarial sample while maintaining two important properties of the feature space, namely statistical correlation and semantic dependency . Although being demonstrated in an OSN security scenario, such an approach might be applied in other context where the aim is to perturb data described by mutually related features. Experiments conducted on a public dataset show the effectiveness of AdverSPAM compared to five state-of-the-art competitors, even in the presence of adversarial defense mechanisms.

Improved Binary Meerkat Optimization Algorithm for efficient feature selection of supervised learning classification

Feature selection (FS) is a crucial step in machine learning and data mining projects. It aims to remove redundant and uncorrelated features, thus improving the accuracy of models. However, it can be challenging to select the optimal features, especially for datasets with many features. This paper proposes the Binary Meerkat Optimization Algorithm (BMOA) to address the issue of FS. The BMOA selects the most related and useful features. Additionally, an improved BMOA (IBMOA) is introduced, which enhances exploration and exploitation capabilities by incorporating the Periodic Mode Boundary Handling (PMBH) strategy and the Local Search (LS) process. This helps to reduce dimensionality and enhance classification accuracy. To evaluate the significance of selected features, two widely used classifiers, namely k-nearest Neighbor (k-NN) and Support Vector Machine (SVM), were used as quality raters. The proposed IBMOA and the original BMOA were compared on 21 multi-scale and multi-faceted benchmark datasets. The binary structures of eight contemporary algorithms, including Binary Artificial Bee Colony (BABC), Binary Grey Wolf Optimization (BGWO), Binary Sailfish Optimizer (BSFO), Binary Particle Swarm Optimizer (BPSO), Binary Bird Swarm Algorithm (BBSA), Binary Whale Optimization Algorithm (BWOA), Binary Grasshopper Optimization Algorithm (BGOA), and Binary Bat Algorithm (BBA), were also analyzed and compared. The proposed IBMOA algorithm outperforms competitors on small and large datasets, obtaining optimal solutions within a reasonable timeframe. This is true for the proposed IBMOA, which has been statistically proven to be highly competitive using the Wilcoxon rank sum test (with alpha=0.05).

Improving soybean yield prediction by integrating UAV nadir and cross-circling oblique imaging

High-throughput estimation of soybean yield using unmanned aerial vehicle (UAV) imagery can help improve the efficiency of soybean breeding. Previous studies have mainly focused on the extraction of vegetation indices and texture features from two-dimensional(2D) orthophotos to construct empirical models of yield, lacking spatial structure information of crops. Therefore, UAV cross-circling oblique (CCO) photography combined with SfM-MVS algorithm was used to reconstruct three-dimensional(3D) soybean canopy structure. Then canopy 3D related phenotypic features are extracted and combined with features from RGB nadir and multispectral images to analyze the capability of different modal data fusion on soybean yield prediction. In addition, Shapley value was used to evaluate the importance of features across different machine learning models. Based on the Shapley value, a bagging-stacking ensemble learning framework was developed using Lasso, Random Forest (RF), Ridge Regression (RR), and XGBoost as base learners for yield prediction. The performance of the traditional stacking method was evaluated and compared with weighted average methods such as Bayesian Model Averaging (BMA) and Entropy Weighted Average (EWA) as meta-learners in the ensemble framework. The results show that 3D canopy structure of soybean can be obtained from UAV CCO photography. The inclusion of 3D structural features can improve the accuracy of yield estimation. Among different modal data combinations, the highest estimation accuracy was achieved when combining RGB nadir features with CCO 3D features. The performance of the above base learners was improved by 8.8%, 3.5%, 7.1%, and 8.0% respectively when using the Shapley value method. The accuracy of yield prediction applied on independent dataset of year 2023 was further calculated by using the bagging-stacking ensemble learning method. When using BMA as the meta-learner, the best performance is obtained with an R2 of 0.7. Therefore, UAV CCO photography with SfM-MVS algorithm provides a new approach to obtain high-quality point clouds of the crop canopy at low cost. UAV multimodal data combined with ensemble learning models allow accurate estimation of yield prediction in breeding materials of soybeans at the plot scale.