General Data Research Articles

Clinical Results of Asymmetric Blepharoplasty Plastic Repair.

To investigate plastic surgery repair's effects and adverse reactions in the clinical therapy of asymmetric double eyelids. All 126 sufferers who came to the hospital for asymmetric double eyelid plastic repair from January 2022 to October 2022 were selected as the research objects, and they were divided into a control group and an observation group using the random number method, with 63 cases in each group, in which sufferers in the control group underwent full incision blepharoplasty and sufferers in the observation group underwent small incision liposuction with submerged sutures. The general data, treatment results, treatment satisfaction, related surgical indicators, and frequency of adverse reactions of the 2 groups of sufferers with asymmetric double eyelid plastic repair were compared. It had no statistically obvious distinction between the control group and the observation group of sufferers in terms of general data such as sex, age, weight, and height (P>0.05); the total therapy efficiency of the sufferers in the observation group (95.24%) was greater than the control group (74.60%), with P value <0.05; the total treatment satisfaction of the patients in the observation group (96.83%) was significantly higher than that in the control group (76.19%), with P value <0.05; compared with the control group, patients in the observation group had shorter operative time and healing time and less intraoperative bleeding, with P value <0.05; the total frequency of adverse reactions of sufferers in the observation group (4.76%) was less than the control group (17.46%), with P value <0.05. Small incision liposuction with submerged sutures for plastic repair has significant efficacy, relatively high patient satisfaction, and low incidence of adverse reactions, in line with patient esthetic review, which has a broad clinical application prospect.

Artificial intelligence based data curation: enabling a patient-centric European health data space.

The emerging European Health Data Space (EHDS) Regulation opens new prospects for large-scale sharing and re-use of health data. Yet, the proposed regulation suffers from two important limitations: it is designed to benefit the whole population with limited consideration for individuals, and the generation of secondary datasets from heterogeneous, unlinked patient data will remain burdensome. AIDAVA, a Horizon Europe project that started in September 2022, proposes to address both shortcomings by providing patients with an AI-based virtual assistant that maximises automation in the integration and transformation of their health data into an interoperable, longitudinal health record. This personal record can then be used to inform patient-related decisions at the point of care, whether this is the usual point of care or a possible cross-border point of care. The personal record can also be used to generate population datasets for research and policymaking. The proposed solution will enable a much-needed paradigm shift in health data management, implementing a 'curate once at patient level, use many times' approach, primarily for the benefit of patients and their care providers, but also for more efficient generation of high-quality secondary datasets. After 15 months, the project shows promising preliminary results in achieving automation in the integration and transformation of heterogeneous data of each individual patient, once the content of the data sources managed by the data holders has been formally described. Additionally, the conceptualization phase of the project identified a set of recommendations for the development of a patient-centric EHDS, significantly facilitating the generation of data for secondary use.

Machine Learning Assisted Exploration for Affine Deligne–Lusztig Varieties

AbstractThis paper presents a novel, interdisciplinary study that leverages a Machine Learning (ML) assisted framework to explore the geometry of affine Deligne–Lusztig varieties (ADLV). The primary objective is to investigate the non-emptiness pattern, dimension, and enumeration of irreducible components of ADLV. Our proposed framework demonstrates a recursive pipeline of data generation, model training, pattern analysis, and human examination, presenting an intricate interplay between ML and pure mathematical research. Notably, our data-generation process is nuanced, emphasizing the selection of meaningful subsets and appropriate feature sets. We demonstrate that this framework has a potential to accelerate pure mathematical research, leading to the discovery of new conjectures and promising research directions that could otherwise take significant time to uncover. We rediscover the virtual dimension formula and provide a full mathematical proof of a newly identified problem concerning a certain lower bound of dimension. Furthermore, we extend an open invitation to the readers by providing the source code for computing ADLV and the ML models, promoting further explorations. This paper concludes by sharing valuable experiences and highlighting lessons learned from this collaboration.

Shape completion in the dark: completing vertebrae morphology from 3D ultrasound.

Ultrasound (US) imaging, while advantageous for its radiation-free nature, is challenging to interpret due to only partially visible organs and a lack of complete 3D information. While performing US-based diagnosis or investigation, medical professionals therefore create a mental map of the 3D anatomy. In this work, we aim to replicate this process and enhance the visual representation of anatomical structures. We introduce a point cloud-based probabilistic deep learning (DL) method to complete occluded anatomical structures through 3D shape completion and choose US-based spine examinations as our application. To enable training, we generate synthetic 3D representations of partially occluded spinal views by mimicking US physics and accounting for inherent artifacts. The proposed model performs consistently on synthetic and patient data, with mean and median differences of 2.02 and 0.03 in Chamfer Distance (CD), respectively. Our ablation study demonstrates the importance of US physics-based data generation, reflected in the large mean and median difference of 11.8 CD and 9.55 CD, respectively. Additionally, we demonstrate that anatomical landmarks, such as the spinous process (with reconstruction CD of 4.73) and the facet joints (mean distance to ground truth (GT) of 4.96 mm), are preserved in the 3D completion. Our work establishes the feasibility of 3D shape completion for lumbar vertebrae, ensuring the preservation of level-wise characteristics and successful generalization from synthetic to real data. The incorporation of US physics contributes to more accurate patient data completions. Notably, our method preserves essential anatomical landmarks and reconstructs crucial injections sites at their correct locations.

Guided Docking as a Data Generation Approach Facilitates Structure-Based Machine Learning on Kinases.

Drug discovery pipelines nowadays rely on machine learning models to explore and evaluate large chemical spaces. While including 3D structural information is considered beneficial, structural models are hindered by the availability of protein-ligand complex structures. Exemplified for kinase drug discovery, we address this issue by generating kinase-ligand complex data using template docking for the kinase compound subset of available ChEMBL assay data. To evaluate the benefit of the created complex data, we use it to train a structure-based E(3)-invariant graph neural network. Our evaluation shows that binding affinities can be predicted with significantly higher precision by models that take synthetic binding poses into account compared to ligand- or drug-target interaction models alone.

Development of a multiplex real-time quantitative reverse-transcription polymerase chain reaction for the detection of four bee viruses

Viruses in the families Dicistroviridae and Iflaviridae are among the main threats to western honey bees (Apis mellifera) and native bee species. Polymerase chain reaction (PCR) is the gold standard for pathogen detection in bees. However, high throughput screening for bee virus infections in singleplex PCR reactions is cumbersome and limited by the high quantities of sample RNA required. Thus, the development of a sensitive and specific multiplex PCR detection method for screening for multiple viruses simultaneously is necessary. Here, we report the development of a one-step multiplex reverse-transcription quantitative polymerase chain reaction (RT-qPCR) assay to detect four viruses commonly encountered in pollinator species. The optimized multiplex RT-qPCR protocol described in this study allows simultaneous detection of two dicistroviruses (Israeli acute paralysis virus and Black queen cell virus) and two iflaviruses (Sacbrood virus and Deformed wing virus) with high efficiency and specificity comparable to singleplex detection assays. This assay provides a broad range of detection and quantification, and the results of virus quantification in this study are similar to those performed in other studies using singleplex detection assays. This method will be particularly useful for data generation from small-bodied insect species that yield low amounts of RNA.

DS-PACK: Tool assembly for the end-to-end support of controlled access human data sharing

The EU General Data Protection Regulation (GDPR) requirements have prompted a shift from centralised controlled access genome-phenome archives to federated models for sharing sensitive human data. In a data-sharing federation, a central node facilitates data discovery; meanwhile, distributed nodes are responsible for handling data access requests, concluding agreements with data users and providing secure access to the data. Research institutions that want to become part of such federations often lack the resources to set up the required controlled access processes. The DS-PACK tool assembly is a reusable, open-source middleware solution that semi-automates controlled access processes end-to-end, from data submission to access. Data protection principles are engraved into all components of the DS-PACK assembly. DS-PACK centralises access control management and distributes access control enforcement with support for data access via cloud-based applications. DS-PACK is in production use at the ELIXIR Luxembourg data hosting platform, combined with an operational model including legal facilitation and data stewardship.

An in-depth evaluation of federated learning on biomedical natural language processing for information extraction

Language models (LMs) such as BERT and GPT have revolutionized natural language processing (NLP). However, the medical field faces challenges in training LMs due to limited data access and privacy constraints imposed by regulations like the Health Insurance Portability and Accountability Act (HIPPA) and the General Data Protection Regulation (GDPR). Federated learning (FL) offers a decentralized solution that enables collaborative learning while ensuring data privacy. In this study, we evaluated FL on 2 biomedical NLP tasks encompassing 8 corpora using 6 LMs. Our results show that: (1) FL models consistently outperformed models trained on individual clients’ data and sometimes performed comparably with models trained with polled data; (2) with the fixed number of total data, FL models training with more clients produced inferior performance but pre-trained transformer-based models exhibited great resilience. (3) FL models significantly outperformed pre-trained LLMs with few-shot prompting.

Blade fouling fault detection based on shaft orbit generative adversarial network

To address the challenges of accuracy and interpretability in mechanical fault detection models, this study proposes a shaft orbit generative adversarial network (SOGAN) and applies it to detect blade fouling faults. Variational autoencoder (VAE) is used as the foundational network architecture for extracting high-dimensional latent features from the shaft orbit images. Concurrently, the invariant moments of the shaft orbit images are extracted and embedded in a bypass within the generator, thereby enhancing the accuracy of fault detection. Two sets of real-world blade fouling fault data are collected and meticulously analyzed. The proposed SOGAN model demonstrates significant performance improvements, with average increases of 18.91%, 10.20%, and 26.79% in accuracy compared to the autoencoder, VAE, and GANomaly algorithms, respectively. The F1 scores for both the groups exceed 0.98. The data generated by the proposed SOGAN model exhibit a trend-wise correspondence with the finite element modeling data. In addition, the use of gradient information for the localization and visual analysis of anomalies dynamically tracks the spatial evolution of the rotor shaft orbit throughout its lifecycle. The data generation capability and interpretability of the proposed model can effectively support digital twin modeling and health management of rotating machinery.

Productivity susceptibility analysis in extremely data-poor scenarios: The case of Peruvian coastal groundfish

Coastal small-scale fisheries remain largely unassessed and unregulated worldwide, which leads to a high risk of overfishing. Availability and quality of data are major barriers to performing model-based assessments, particularly in low and middle-income countries. In that context, alternative assessment methods are key to defining priorities, as well as prompting and guiding management actions. In this study, we assessed the vulnerability to overfishing of 10 extremely data-poor species of coastal groundfish in Peru, by applying an adapted Productivity Susceptibility Analysis (PSA). We classified fisheries-related Susceptibility attributes and streamlined a process to assess and integrate multiple fisheries and gears into a single score for each attribute. The data quality scores regarding productivity and susceptibility were between “limited” and “no data”, across all stocks. Therefore, expert knowledge was key for completing the assessment. Stocks of four species presented extremely high Vulnerability to overfishing: broomtail grouper (V=2.57), grape-eye seabass (V=2.50), pacific goliath grouper (V=2.28) and galapagos sheephead wrasse (V=2.24). Three showed high Vulnerability to overfishing: black snook (V=2.13), harlequin wrasse (V=2.05) and chino (V=2.01). Two showed medium Vulnerability to overfishing: mulata (V=1.89) and pacific beakfish (V=1.89). Only the stock of bumphead parrotfish showed low Vulnerability to overfishing (V=1.73). We contrasted the resulting Vulnerability scores with the limited available information on the statuses of these stocks, to evaluate the accurateness of the overfishing probability thresholds proposed in PSA literature. We found that Vulnerability scores were mostly consistent with contrasted information and that the thresholds remain useful to inform fisheries management in extremely data-poor contexts. Our results shed light on the need for the implementation of fishing regulations and the generation of biological and fishery-related data on coastal groundfish in Peru.

A Balancing Act: Data Protection Compliance of Artificial Intelligence

Abstract Neural network based generative artificial intelligence (GenAI) systems, in particular large language models (LLMs), hold huge potential for economic, scientific, and artistic progress. However, they are also the subject of ongoing fierce data protection debates. Some even doubt whether GenAI systems can, at least in principle, comply with the General Data Protection Regulation (EU) 2016/679 (GDPR). This issue was highlighted by a temporary ban of ChatGPT by the Italian data protection authority in March 2023. To answer the question of possible compliance with the GDPR, this paper will first illustrate relevant technical aspects and then examine the extent to which GenAI systems process personal data. A distinction will be made between training, use, and the system itself as stored on a hard drive. On this basis, this paper addresses the key question of the lawfulness of the respective processing activities, bearing in mind that each specific use case of any GenAI system requires a specific data protection assessment. While the issues discussed in this paper are relevant to GenAI systems in general, they are particularly relevant to LLMs trained on large amounts of text-based personal data. As one result, the analysis identifies serious challenges in the application of the GDPR to LLMs. However, with no significant updates to the relevant GDPR provisions in sight, this paper aims to show how pragmatic solutions are both necessary and largely possible even under the current legal framework. This includes practical guidance for developers and users on how to ensure compliance.

The relationship between serum coagulation parameters and the recurrence of chronic subdural hematoma.

The aim is to investigate the relationship between serum coagulation parameters (PT, APTT, D-D and FDP) before hospitalization and recurrence of chronic subdural hematoma (CSDH). 236 patients with CSDH who were diagnosed for the first time and had complete medical records were followed up for at least 90days. Fifty patients (21.2%) had relapsed. Univariate analysis was conducted including general data, imaging data and test results. Serum coagulation parameters (PT, APTT, D-D and FDP) were detected for all CSDH patients. The study identified several factors that exhibited a significant correlation with chronic subdural hematoma (CSDH) recurrence. These factors included advanced age (p = 0.01), hypertension (p = 0.04), liver disease (p = 0.01), anticoagulant drug use (p = 0.01), antiplatelet drug use (p = 0.02), bilateral hematoma (p = 0.02), and single-layer hematoma (p = 0.01). In addition, the presence of fibrin/fibrinogen degradation products (FDP) exceeding 5mg/L demonstrated a significant relationship with CSDH recurrence (P < 0.05). Notably, the combined assessment of D-dimer (D-D) and FDP exhibited a significant difference, particularly regarding recurrence within 30days after surgery (P < 0.05). The simultaneous elevation of serum FDP and D-D levels upon admission represents a potentially novel predictor for CSDH recurrence. This finding is particularly relevant for patients who experience recurrence within 30days following surgical intervention. Older individuals with CSDH who undergo trepanation and drainage should be closely monitored due to their relatively higher recurrence rate.

Will the EU AI Regulations Give Rise to Another ‘Brussels Effect’? Lessons from the GDPR

The pre-eminence of the European Union (EU) General Data Protection Regulation (GDPR) in regulating the business of data collection, processing and transfer cannot be understated. It has come to serve as the model for laws in several non-EU jurisdictions who share in the EU’s concerns about citizens’ data being harnessed by Big Tech in particular. This article explores the GDPR’s outsized impact in the sphere of data regulation, the conflict between the contrasting models of regulation adopted by the EU and the United States and comments on the possibility of the AI regulations becoming what the GDPR is for data. It further delves into the possibility of this approach being followed by countries like India, where government efforts to integrate AI-based innovation and entrepreneurship have led to the possibility of newer regulatory approaches being adopted.

Views of general practice staff on sharing general practice data for research: a scoping review protocol.

The objective of this scoping review is to understand the range and types of evidence in relation to the views of general practitioner and other general practice staff on sharing general practice data for research purposes. The use of general practice data for research has the potential to drive transformative improvements in health care. The vast amount of patient data collected in general practice is valuable and provides researchers with data to conduct large-scale studies and generate evidence that can inform policy decisions, support the development of personalized medicine, and enhance patient outcomes. However, despite there being clear benefits to using general practice data for research, there are also potential harms, such as data misuse, loss of trust between the general practitioner and patient, and data breaches. This scoping review will focus on the views of general practice staff, including general practitioners, practice nurses, and practice managers, about sharing general practice data for the purposes of research. This scoping review will exclude sources of evidence that are conducted outside of the general practice setting, and papers that report on public, patient, or community views on data sharing. This scoping review will be conducted in accordance with JBI methodology for scoping reviews. A 3-step search strategy will be used to acquire both published and unpublished sources of evidence. Two independent reviewers will select sources of evidence in line with the inclusion and exclusion criteria. No limits on the date of the search or language will be applied. Data will be extracted and the results will be summarized descriptively and presented in a tabular format. Open Science Framework https://osf.io/49yw5.

CrossHAR: Generalizing Cross-dataset Human Activity Recognition via Hierarchical Self-Supervised Pretraining

The increasing availability of low-cost wearable devices and smartphones has significantly advanced the field of sensor-based human activity recognition (HAR), attracting considerable research interest. One of the major challenges in HAR is the domain shift problem in cross-dataset activity recognition, which occurs due to variations in users, device types, and sensor placements between the source dataset and the target dataset. Although domain adaptation methods have shown promise, they typically require access to the target dataset during the training process, which might not be practical in some scenarios. To address these issues, we introduce CrossHAR, a new HAR model designed to improve model performance on unseen target datasets. CrossHAR involves three main steps: (i) CrossHAR explores the sensor data generation principle to diversify the data distribution and augment the raw sensor data. (ii) CrossHAR then employs a hierarchical self-supervised pretraining approach with the augmented data to develop a generalizable representation. (iii) Finally, CrossHAR fine-tunes the pretrained model with a small set of labeled data in the source dataset, enhancing its performance in cross-dataset HAR. Our extensive experiments across multiple real-world HAR datasets demonstrate that CrossHAR outperforms current state-of-the-art methods by 10.83% in accuracy, demonstrating its effectiveness in generalizing to unseen target datasets.

Generative Adversarial Networks for Synthetic Data Generation in Finance: Evaluating Statistical Similarities and Quality Assessment

Generating synthetic data is a complex task that necessitates accurately replicating the statistical and mathematical properties of the original data elements. In sectors such as finance, utilizing and disseminating real data for research or model development can pose substantial privacy risks owing to the inclusion of sensitive information. Additionally, authentic data may be scarce, particularly in specialized domains where acquiring ample, varied, and high-quality data is difficult or costly. This scarcity or limited data availability can limit the training and testing of machine-learning models. In this paper, we address this challenge. In particular, our task is to synthesize a dataset with similar properties to an input dataset about the stock market. The input dataset is anonymized and consists of very few columns and rows, contains many inconsistencies, such as missing rows and duplicates, and its values are not normalized, scaled, or balanced. We explore the utilization of generative adversarial networks, a deep-learning technique, to generate synthetic data and evaluate its quality compared to the input stock dataset. Our innovation involves generating artificial datasets that mimic the statistical properties of the input elements without revealing complete information. For example, synthetic datasets can capture the distribution of stock prices, trading volumes, and market trends observed in the original dataset. The generated datasets cover a wider range of scenarios and variations, enabling researchers and practitioners to explore different market conditions and investment strategies. This diversity can enhance the robustness and generalization of machine-learning models. We evaluate our synthetic data in terms of the mean, similarities, and correlations.

Enhancing Workplace Safety through Personalized Environmental Risk Assessment: An AI-Driven Approach in Industry 5.0

This paper describes an integrated monitoring system designed for individualized environmental risk assessment and management in the workplace. The system incorporates monitoring devices that measure dust, noise, ultraviolet radiation, illuminance, temperature, humidity, and flammable gases. Comprising monitoring devices, a server-based web application for employers, and a mobile application for workers, the system integrates the registration of workers’ health histories, such as common diseases and symptoms related to the monitored agents, and a web-based recommendation system. The recommendation system application uses classifiers to decide the risk/no risk per sensor and crosses this information with fixed rules to define recommendations. The system generates actionable alerts for companies to improve decision-making regarding professional activities and long-term safety planning by analyzing health information through fixed rules and exposure data through machine learning algorithms. As the system must handle sensitive data, data privacy is addressed in communication and data storage. The study provides test results that evaluate the performance of different machine learning models in building an effective recommendation system. Since it was not possible to find public datasets with all the sensor data needed to train artificial intelligence models, it was necessary to build a data generator for this work. By proposing an approach that focuses on individualized environmental risk assessment and management, considering workers’ health histories, this work is expected to contribute to enhancing occupational safety through computational technologies in the Industry 5.0 approach.

A Causality-Aware Perspective on Domain Generalization via Domain Intervention

Most mainstream statistical models will achieve poor performance in Out-Of-Distribution (OOD) generalization. This is because these models tend to learn the spurious correlation between data and will collapse when the domain shift exists. If we want artificial intelligence (AI) to make great strides in real life, the current focus needs to be shifted to the OOD problem of deep learning models to explore the generalization ability under unknown environments. Domain generalization (DG) focusing on OOD generalization is proposed, which is able to transfer the knowledge extracted from multiple source domains to the unseen target domain. We are inspired by intuitive thinking about human intelligence relying on causality. Unlike relying on plain probability correlations, we apply a novel causal perspective to DG, which can improve the OOD generalization ability of the trained model by mining the invariant causal mechanism. Firstly, we construct the inclusive causal graph for most DG tasks through stepwise causal analysis based on the data generation process in the natural environment and introduce the reasonable Structural Causal Model (SCM). Secondly, based on counterfactual inference, causal semantic representation learning with domain intervention (CSRDN) is proposed to train a robust model. In this regard, we generate counterfactual representations for different domain interventions, which can help the model learn causal semantics and develop generalization capacity. At the same time, we seek the Pareto optimal solution in the optimization process based on the loss function to obtain a more advanced training model. Extensive experimental results of Rotated MNIST and PACS as well as VLCS datasets verify the effectiveness of the proposed CSRDN. The proposed method can integrate causal inference into domain generalization by enhancing interpretability and applicability and brings a boost to challenging OOD generalization problems.

The right to be forgotten: separate conceptual issues

The article is devoted to the study of the emergence and formation of the human right to be forgotten as the right of an individual to request the removal of his or her personal data from public access through search engines, under certain conditions. It is found that the establishment of the right to be forgotten is often associated with one of the key decisions of the Court of Justice of the European Union, namely, in the case of Google Spain v. Costeja Gonzalez, which has gained wide international publicity. However, it is important to note that freedom of speech and freedom of access to information are conflicting fundamental rights to the right to be forgotten. The right to privacy and the data protection regime do not exist in a vacuum: the right to privacy can often clash with other rights, in particular the right to freedom of expression, for example, in terms of censorship. The author analyses international judicial precedents regarding the acknowledgement of the right to be forgotten in the sphere of the digital human rights system, and finds that the rulings typically pertain to data that might drastically influence the fundamental rights to privacy and personal data protection. Furthermore, the courts explicitly assert their intention to achieve an equitable balance between these rights and the public's right to access information. Hence, the resolution of numerous conflicts is largely contingent on the factual context of each case. The author reviews the legal regulation of the right to be forgotten, in particular, the Regulation of the European Parliament and of the Council of 27 April 2016 (General Data Protection Regulation), the 1995 Data Protection Directive, as updated and supplemented by the above-mentioned Regulation, and the European Convention on Human Rights, etc. Although the concept of the right to be forgotten is not recognised in the Ukrainian legal system, the Law of Ukraine "On Personal Data Protection” states that deletion or destruction of personal data in the information landscape is possible upon filing a complaint or going to court.

Predicting mixed venous oxygen saturation (SvO2) impairment in COPD patients using clinical-CT radiomics data: A preliminary study.

Chronic obstructive pulmonary disease (COPD) is one of the most common chronic airway diseases in the world. To predict the degree of mixed venous oxygen saturation (SvO2) impairment in patients with COPD by modeling using clinical-CT radiomics data and to provide reference for clinical decision-making. A total of 236 patients with COPD diagnosed by CT and clinical data at Xiangyang No. 1 People's Hospital (n= 157) and Xiangyang Central Hospital (n= 79) from June 2018 to September 2021 were retrospectively analyzed. The patients were divided into group A (SvO⩾2 62%, N= 107) and group B (SvO<2 62%, N= 129). We set up training set and test set at a ratio of 7/3 and time cutoff spot; In training set, Logistic regression was conducted to analyze the differences in general data (e.g. height, weight, systolic blood pressure), laboratory indicators (e.g. arterial oxygen saturation and pulmonary artery systolic pressure), and CT radiomics (radscore generated using chest CT texture parameters from 3D slicer software and LASSO regression) between these two groups. Further the risk factors screened by the above method were used to establish models for predicting the degree of hypoxia in COPD, conduct verification in test set and create a nomogram. Univariate analysis demonstrated that age, smoking history, drinking history, systemic systolic pressure, digestive symptoms, right ventricular diameter (RV), mean systolic pulmonary artery pressure (sPAP), cardiac index (CI), pulmonary vascular resistance (PVR), 6-min walking distance (6MWD), WHO functional classification of pulmonary hypertension (WHOPHFC), the ratio of forced expiratory volume in the first second to the forced vital capacity (FEV1%), and radscore in group B were all significantly different from those in group A (P< 0.05). Multivariate regression demonstrated that age, smoking history, digestive symptoms, 6MWD, and radscore were independent risk factors for SvO2 impairment. The combined model established based on the abovementioned indicators exhibited a good prediction effect [AUC: 0.903; 95%CI (0.858-0.937)], higher than the general clinical model [AUC: 0.760; 95%CI (0.701-0.813), P< 0.05] and laboratory examination-radiomics model [AUC: 0.868; 95%CI (0.818-0.908), P= 0.012]. The newly created nomogram may be helpful for clinical decision-making and benefit COPD patients. SvO2 is an important indicator of hypoxia in COPD, and it is highly related to age, 6MWD, and radscore. The combined model is helpful for early identification of SvO2 impairment and adjustment of COPD treatment strategies.