Neighbor Clustering Research Articles

Barrier function–based adaptive STSMC for steering feel feedback of SBW vehicle with uncertainty

As an essential component of human–computer interaction in steer-by-wire (SBW) vehicles, the steering feel feedback system provides drivers with accurate steering feel by precisely controlling the output torque of the steering feel motor. Meanwhile, the existence of uncertainties makes the accurate control of steering feel feedback torque become an acknowledged challenging issue. This article proposes an adaptive super-torque sliding mode control strategy based on barrier functions for the steering feel feedback system of SBW vehicles with highly complex dynamics in the presence of uncertainties. First, to achieve intelligent modeling under the influence of uncertainty factors, a dynamic system with unknown uncertainty based on adaptive fuzzy logic system (FLS) is proposed. In the framework of this model, the uncertainty factors introduced by unknown external disturbance, parameter uncertainty, and model couplings are considered as a lumped uncertainty function. Furthermore, the membership functions of the FLS are dynamically adjusted based on nearest neighbor clustering, aiming to enhance the modeling accuracy beyond traditional fuzzy modeling. To further offset the system uncertainty and FLS approximation error, an adaptive super-twisting sliding mode control (STSMC) method based on barrier function is proposed. This algorithm retains the advantages of traditional sliding mode control (SMC) methods in dealing with systems with uncertainty and eliminates the need for a priori knowledge of the upper bound of uncertainties by introducing the adaptive law based on barrier function, avoiding the use of complex identification algorithms. Moreover, it eliminates the gain of overestimation and significantly reduces chattering phenomenon in traditional sliding mode control. Finally, based on Lyapunov stability theory, this article proves the proposed system can achieve convergence within a finite time. Compared with STSMC, traditional SMC and proportional–integral–derivative (PID) control, the results verify the effectiveness of the proposed control method. The proposed strategy provides a new solution to reduce the steering feel feedback torque error under the influence of uncertain factors.

Large-area magnetic skin for multi-point and multi-scale tactile sensing with super-resolution

The advancements in tactile sensor technology have found wide-ranging applications in robotic fields, resulting in remarkable achievements in object manipulation and overall human-machine interactions. However, the widespread availability of high-resolution tactile skins remains limited, due to the challenges of incorporating large-sized, robust sensing units and increased wiring complexity. One approach to achieve high-resolution and robust tactile skins is to integrate a limited number of sensor units (taxels) into a flexible surface material and leverage signal processing techniques to achieve super-resolution sensing. Here, we present a magnetic skin consisting of multi-direction magnetized flexible films and a contactless Hall sensor array. The key features of the proposed sensor include the specific magnetization arrangement, K-Nearest Neighbors (KNN) clustering algorithm and convolutional neural network (CNN) model for signal processing. Using only an array of 4*4 taxels, our magnetic skin is capable of achieving super-resolution perception over an area of 48400 mm2, with an average localization error of 1.2 mm. By employing neural network algorithms to decouple the multi-dimensional signals, the skin can achieve multi-point and multi-scale perception. We also demonstrate the promising potentials of the proposed sensor in intelligent control, by simultaneously controlling two vehicles with trajectory mapping on the magnetic skin.

Perceptual point cloud qality assessment for immersive metaverse experience

Perceptual quality assessment for point cloud is critical for immersive metaverse experience and is a challenging task. Firstly, because point cloud is formed by unstructured 3D points that makes the topology more complex. Second, the quality impairment generally involves both geometric attributes and color properties, where the measurement of the geometric distortion becomes more complex. We proposed a perceptual point cloud quality assessment model that follows the perceptual features of Human Visual System (HVS) and the intrinsic characteristics of the point cloud. The point cloud is first pre-processed to extract the geometric skeleton keypoints with graph filtering-based re-sampling, and local neighboring regions around the geometric skeleton keypoints are constructed by K-Nearest Neighbors (KNN) clustering. For geometric distortion, the Point Feature Histogram (PFH) is extracted as the feature descriptor, and the Earth Mover's Distance (EMD) between the PFHs of the corresponding local neighboring regions in the reference and the distorted point clouds is calculated as the geometric quality measurement. For color distortion, the statistical moments between the corresponding local neighboring regions are computed as the color quality measurement. Finally, the global perceptual quality assessment model is obtained as the linear weighting aggregation of the geometric and color quality measurement. The experimental results on extensive datasets show that the proposed method achieves the leading performance as compared to the state-of-the-art methods with less computing time. Meanwhile, the experimental results also demonstrate the robustness of the proposed method across various distortion types. The source codes are available at https://github.com/llsurreal919/PointCloudQualityAssessment.

SNN-PDM: An Improved Probability Density Machine Algorithm Based on Shared Nearest Neighbors Clustering Technique

SNN-PDM: An Improved Probability Density Machine Algorithm Based on Shared Nearest Neighbors Clustering Technique

Unified multi-level neighbor clustering for Source-Free Unsupervised Domain Adaptation

Source-Free Unsupervised Domain Adaptation (SFUDA) transfers knowledge from the source domain to the target domain with using the source domain model instead of the source data, as the source data cannot be accessed in data privacy scenarios. Our method is based on a clustering assumption: although there is a domain shift, target data with similar semantic still form a cluster in the source feature space. We identify two levels of clustering. One is class-level neighbor clustering: data with the same label tend to form a large cluster. The other is instance-level neighbor clustering: data and its neighbors tend to share the same label. Previous methods only consider one level, and we consider that both are complementary. We propose a new SFUDA method called unified multi-level neighbor clustering to address class and instance consistency in a complementary way. Our proposed method achieves competitive results on three domain adaptation benchmark datasets.

Seeing All From a Few: Nodes Selection Using Graph Pooling for Graph Clustering.

Recently, there has been considerable research interest in graph clustering aimed at data partition using graph information. However, one limitation of most graph-based methods is that they assume that the graph structure to operate is reliable. However, there are inevitably some edges in the graph that are not conducive to graph clustering, which we call spurious edges. This brief is the first attempt to employ the graph pooling technique for node clustering to the best of our knowledge. In this brief, we propose a novel dual graph embedding network (DGEN), which is designed as a two-step graph encoder connected by a graph pooling layer to learn the graph embedding. In DGEN, we assume that if a node and its nearest neighboring node are close to the same clustering center, this node is informative, and this edge can be considered as a cluster-friendly edge. Based on this assumption, the neighbor cluster pooling (NCPool) is devised to select the most informative subset of nodes and the corresponding edges based on the distance of nodes and their nearest neighbors to the cluster centers. This can effectively alleviate the impact of the spurious edges on the clustering. Finally, to obtain the clustering assignment of all nodes, a classifier is trained using the clustering results of the selected nodes. Experiments on five benchmark graph datasets demonstrate the superiority of the proposed method over state-of-the-art algorithms.

Abstract 4927: Combining single-cell ATAC and RNA sequencing for supervised cell annotation

Abstract Background: Analysis of samples at the single-cell level offers insights into cellular heterogeneity and cell function. Cell type annotation is the first critical step for performing such an analysis. While current methods primarily utilize single-cell RNA sequencing (scRNA-seq) for annotation, several studies have demonstrated improved classification accuracy by combining scRNA-seq with transposase-accessible chromatin sequencing (ATAC-seq) using unsupervised methods. However, the utility of ATAC-seq features for supervised cell-type annotation has not been explored. Aims/Objectives: The objective of this study was to evaluate the relative performance of supervised cell-type classification using scRNA-seq alone vs. in multimodal combination with ATAC-seq; and how these data interplay with choice of classification and dimensionality reduction methods. Methods: A peripheral-blood mononuclear cell multi-omic dataset from a single, healthy female donor wasanalysed in this study. Ground truth annotations were generated using unsupervised annotation with the weighted nearest neighbour clustering method. Two dimensionality reduction methods (principal component analysis (PCA), single-cell Variational Inference (scVI) autoencoder) and four classification models (logistic regression, random forest, support vector machine (SVM)) were implemented and performance metrics (F1 score, precision, and recall) were compared over 10 bootstrap samples. Results: ATAC-seq features improved annotation quality and prediction confidence when using scVI embeddings, independent of the classifier. The best-performing model (SVM with scVI embeddings) showed an increase from a median macro F1 score of 0.907 (IQR = [0.902, 0.910]) using scRNA-seq alone to 0.946 (IQR = [0.940, 0.949], p <0.05) with ATAC-seq added. For PCA embeddings, improvements in macro F1 score were insignificant. All cell types (B, T, monocytes, natural killer and dendritic cells) showed significant improvements when using ATAC-seq with scVI embeddings. CD4 T effector memory cells showed the largest gain in F1 score (0.112, p <0.01), whilst type-2 conventional dendritic cells showed the smallest improvement (0.006, p <0.05). Prediction confidence was improved in B cells, monocytes, natural killer cells, CD4 and CD8 naïve cells, CD4 T central memory cells and CD8 T effector memory cells. Improvements in F1 scores were lost when only classifying major cell types rather than subtypes. Conclusions: Employing ATAC-seq embeddings with scVI autoencoder enhances supervised annotation quality over scRNA-only methods. Further studies should explore the use of ATAC to improve the annotation of highly heterogeneous tissues such as tumours. Citation Format: Jaidip Gill, Abhijit Dasgupta, Brychan Manry, Natasha Markuzon. Combining single-cell ATAC and RNA sequencing for supervised cell annotation [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 4927.

Bimetallic iron complex constructed clusters and single atoms neighboring structure to enhance oxygen reduction reaction performance

Electrocatalysts are useful in lowering the energy barrier in oxygen reduction reaction (ORR). In this study, a catalyst with neighboring Fe single-atom and cluster is created by adsorbing a bimetallic Fe complex onto N-doped carbon and then pyrolyzing it. The resulting catalyst has good performance and a half-wave potential of 0.89 V. When used in Zn-air batteries, the voltage drops by only 8.13 % after 145 h of cycling. Theoretical studies show that electrons transfer from neighboring clusters to single atoms and the catalyst has a lower d-band center. These reduce intermediate desorption energy, hence improving ORR performance. This work demonstrates the capacity to adjust the catalytic properties through the interaction of diverse metal structures, which helps to design more efficient catalysts.

Adjacent MnOx clusters enhance the hydroformylation activity of rhodium single-atom catalysts

Hydroformylation reactions promoted by Rh single-atom catalysts typically exhibit a negative reaction order for CO and a positive reaction order for H2 and alkenes. To enhance the catalytic activity, efforts have been concentrated on reducing the CO adsorption strength and/or enhancing the hydrogenation capacity at Rh sites. This report introduces an optimized Rh single-atom catalyst, utilizing positively charged Mn species to weaken the CO adsorption strength. Our strategy involves placing MnOx clusters in the proximity to the Rh single atom. This arrangement allows the reduction of Rh3+ to produce electronically deficient Rhδ+ species (1 < δ < 2), rather than the usual formation of the lower valence state Rh+ species. The weakened CO adsorption strength on the more positively charged Rhδ+ results in reduced CO coverage on the Rh active site, and enhanced adsorption of propylene. Our mechanistic study reveals that H2 and CO adsorption on Rh catalyzes the reduction of adjacent Mn(IV) species, inducing the formation of a Rh-MnOx paired active site. The neighboring MnOx clusters hinder the extensive reduction of Rh3+ to Rh+. This study presents MnOx as an inorganic ligand in Rh-MnOx pair site modifies the electronic properties of Rh sites to modulate the hydroformylation activity of Rh single-atom catalysts.

Comparison of the effects of imputation methods for missing data in predictive modelling of cohort study datasets

BackgroundMissing data is frequently an inevitable issue in cohort studies and it can adversely affect the study's findings. We assess the effectiveness of eight frequently utilized statistical and machine learning (ML) imputation methods for dealing with missing data in predictive modelling of cohort study datasets. This evaluation is based on real data and predictive models for cardiovascular disease (CVD) risk.MethodsThe data is from a real-world cohort study in Xinjiang, China. It includes personal information, physical examination data, questionnaires, and laboratory biochemical results from 10,164 subjects with a total of 37 variables. Simple imputation (Simple), regression imputation (Regression), expectation-maximization(EM), multiple imputation (MICE) , K nearest neighbor classification (KNN), clustering imputation (Cluster), random forest (RF), and decision tree (Cart) were the chosen imputation methods. Root Mean Square Error (RMSE) and Mean Absolute Error (MAE) are utilised to assess the performance of different methods for missing data imputation at a missing rate of 20%. The datasets processed with different missing data imputation methods were employed to construct a CVD risk prediction model utilizing the support vector machine (SVM). The predictive performance was then compared using the area under the curve (AUC).ResultsThe most effective imputation results were attained by KNN (MAE: 0.2032, RMSE: 0.7438, AUC: 0.730, CI: 0.719-0.741) and RF (MAE: 0.3944, RMSE: 1.4866, AUC: 0.777, CI: 0.769-0.785). The subsequent best performances were achieved by EM, Cart, and MICE, while Simple, Regression, and Cluster attained the worst performances. The CVD risk prediction model was constructed using the complete data (AUC:0.804, CI:0.796-0.812) in comparison with all other models with p<0.05.ConclusionKNN and RF exhibit superior performance and are more adept at imputing missing data in predictive modelling of cohort study datasets.

Assessment of the urban habitat quality service functions and their drivers based on the fusion module of graph attention network and residual network

ABSTRACT Land use/cover change is a major cause of ecological degradation. Reliable LUCC data are essential for evaluating habitat quality. The current method of surface cover classification based on the convolutional neural networks (CNNs) is usually a local spatial operation using a regular convolutional kernel, which ignores the correlation between adjacent image elements. This paper proposes a combination network with two branches, branch 1 uses the K-nearest neighbor clustering algorithm to construct superpixels and then uses the data transformation module to construct a graph attention network (GAT); branch 2 constructs the CNN using attention and residual modules to obtain the spatial and higher-order semantic information of the images. Finally, the features are fused using weighted fusion, and a classification map with less point noise and greater consistency with the real surface coverage is obtained. The classification results of this network are better than those of the other competitive methods. In addition, the urbanization of Sanya has resulted in significant habitat degradation. A good fit ( R 2 in 2020 = 0.639) between habitat quality (HQ) and natural and socioeconomic factors was observed in Sanya. Natural factors are more relevant to HQ than socioeconomic factors and vary spatially.

Segmentation of Coronary Arteries from X-ray Angiographic Images Using a Combination of K-Nearest Neighbor Clustering and Morphological Reconstruction Techniques

Segmentation of Coronary Arteries from X-ray Angiographic Images Using a Combination of K-Nearest Neighbor Clustering and Morphological Reconstruction Techniques

Exploring the Expression and Function of T Cell Surface Markers Identified through Cellular Indexing of Transcriptomes and Epitopes by Sequencing.

By utilizing both protein and mRNA expression patterns, we can identify more detailed and diverse immune cells, providing insights into understanding the complex immune landscape in cancer ecosystems. This study was performed by obtaining publicly available Cellular Indexing of Transcriptomes and Epitopes by Sequencing (CITE-seq) data of peripheral blood mononuclear cells (PBMCs) from the Gene Expression Omnibus database. A total of 94674 total cells were analyzed, of which 32412 were T cells. There were 228 protein features and 16262 mRNA features in the data. The Seurat package was used for quality control and preprocessing, principal component analysis was performed, and Uniform Manifold Approximation and Projection was used to visualize the clusters. Protein and mRNA levels in the CITE-seq were analyzed. We observed that a subset of T cells in the clusters generated at the protein level divided better. By identifying mRNA markers that were highly correlated with the CD4 and CD8 proteins and cross-validating CD26 and CD99 markers using flow cytometry, we found that CD4+ and CD8+ T cells were better discriminated in PBMCs. Weighted Nearest Neighbor clustering results identified a previously unobserved T cell subset. In this study, we used CITE-seq data to confirm that protein expression patterns could be used to identify cells more precisely. These findings will improve our understanding of the heterogeneity of immune cells in the future and provide valuable insights into the complexity of the immune response in health and disease.

Public Perception of ChatGPT and Transfer Learning for Tweets Sentiment Analysis Using Wolfram Mathematica

Understanding public opinion on ChatGPT is crucial for recognizing its strengths and areas of concern. By utilizing natural language processing (NLP), this study delves into tweets regarding ChatGPT to determine temporal patterns, content features, and topic modeling and perform a sentiment analysis. Analyzing a dataset of 500,000 tweets, our research shifts from conventional data science tools like Python and R to exploit Wolfram Mathematica’s robust capabilities. Additionally, with the aim of solving the problem of ignoring semantic information in the LDA model feature extraction, a synergistic methodology entwining LDA, GloVe embeddings, and K-Nearest Neighbors (KNN) clustering is proposed to categorize topics within ChatGPT-related tweets. This comprehensive strategy ensures semantic, syntactic, and topical congruence within classified groups by utilizing the strengths of probabilistic modeling, semantic embeddings, and similarity-based clustering. While built-in sentiment classifiers often fall short in accuracy, we introduce four transfer learning techniques from the Wolfram Neural Net Repository to address this gap. Two of these techniques involve transferring static word embeddings, “GloVe” and “ConceptNet”, which are further processed using an LSTM layer. The remaining techniques center on fine-tuning pre-trained models using scantily annotated data; one refines embeddings from language models (ELMo), while the other fine-tunes bidirectional encoder representations from transformers (BERT). Our experiments on the dataset underscore the effectiveness of the four methods for the sentiment analysis of tweets. This investigation augments our comprehension of user sentiment towards ChatGPT and emphasizes the continued significance of exploration in this domain. Furthermore, this work serves as a pivotal reference for scholars who are accustomed to using Wolfram Mathematica in other research domains, aiding their efforts in text analytics on social media platforms.

Abstract 16780: Single Cell Profiles of Human Heart Failure and Myocardial Diseases

Background: Heart failure represents the end-stage of a range of diseases which adversely impact the heart muscle. The extent by which cellular and molecular mechanisms differ between myocardial diseases in heart failure remains incompletely understood. Methods: We performed isolation and droplet-based RNA-sequencing of 394 247 single nuclei from 103 explanted hearts from transplant and ventricular assist device recipients and 7 unused donor hearts without heart disease. After sample and nuclei level quality control, doublet removal and sample integration; principal component analysis and shared nearest neighbor clustering were used to identify distinct cell populations. Cell type composition and transcriptomic profiles were compared across donor hearts and 10 myocardial diseases (hypertrophic, dilated, ischemic, restrictive, arrythmogenic and non-compaction cardiomyopathy; rheumatic and congenital heart disease; cardiac amyloidosis and graft rejection; n=1-45/group). Results: We identified and manually annotated 14 distinct cell populations that were present across controls and myocardial diseases, including transcriptionally diverse subgroups of cardiomyocytes, endothelial cells and pericytes. Compared to controls, we observed disease-specific differences in cell populations with regards to composition, heterogeneity, and specific transcriptional profiles. Conclusions: Our results reveal differences in single cell profiles of individual heart muscle diseases compared to controls without heart disease.

AN OVERVIEW OF EXISTING LITERATURE ON DOCUMENT SKEW DETECTION

This article presents somewhat detail discussions on document image skew detection methods available in the existing literature. We first categorise the document images into several types and then we discuss different methods available to detect the skew angle of document images belonging to the individual categories. Various methods of document image skew detection include Hough Transform, Principal Components Analysis, Projection Profile based methods, Nearest Neighbor clustering, Connected Component Analysis, Cross Correlation, Radon Transform etc. Discussions on different standard databases of document image samples that have been used by the authors for experimentation purpose are included in the survey. These databases include NIST Federal Register Document Image Database, UW-I, UW-II, UW-III databases, ICDAR DISEC 2013 database, IFN/ENIT database etc. Here, we consider about eight such standard databases. Moreover, we compare different methods with respect to their accuracy and the respective range of skew angles detected by the concerned methods.

Crystal structure of Cs2GdNb6Cl15O3 in the structural evolution of niobium oxychlorides with octa-hedral Nb6-cluster units.

Cs2GdNb6Cl15O3, dicaesium gadolinium hexa-niobium penta-deca-chloride trioxide, was synthesized by solid-state reactions starting from a stoichiometric mixture of CsCl, Gd2O3, Nb, NbCl5, and Nb2O5. The crystal structure is based on octa-hedral Nb6 cluster units (point group symmetry 3.2) with composition [(Nb6Cli 9Oi 3)Cla 6]5- where i and a denote inner and outer ligands. Cs2GdNb6Cl15O3 exhibits 14 valence electrons per cluster unit. The cluster units are linked to each other by CsI and GdIII atoms, whereby CsI (site symmetry 3..) is 12-coordinated by six Cli and six Cla ligands belonging to six neighboring cluster units and GdIII (site symmetry 3.2) is 9-coordinated by three Oi and six Cli ligands belonging to three adjacent cluster units. The arrangement of cluster units corresponds to a stacking of …AA'A… layers along [001]. Cs2GdNb6Cl15O3 is isotypic with Cs2UNb6Cl15O3.

High-Precision Matching Algorithm for Multi-Image Segmentation of Micro Animation Videos in Mobile Network Environment

In the mobile network environment, the accuracy of related image matching algorithms is affected by factors such as bandwidth uncertainty and channel interference, resulting in significant limitations in image feature matching. This article designs a high-precision matching algorithm for multi-image segmentation of micro animation videos in mobile network environments. Fully denoise micro animation video images using 2D High Density Discrete Wavelet Transform (HD-DWT), and apply fixed block count segmentation to process micro animation video images; Using Harris algorithm to complete image corner detection and obtain corner features of sub images; In the K-means clustering algorithm, SIFT feature vectors are divided into clusters and paired with the nearest neighbor cluster in another sub image to form a sub image matching pair, completing block based sub image matching; Combine all sub image matching results to obtain video image matching results, and use the Improved Random Sampling Consistency (RANCAS) algorithm to remove incorrect matching during the matching process, improving matching accuracy. The experimental results show that the designed algorithm can effectively reduce image noise, improve image quality, and generate a large number of matching pairs in mobile network environments. After the application of the designed algorithm, the production effect of micro animated videos in mobile networks can be significantly improved.

A Fully Online Clustering Approach for Enhanced Performance of Health Information System

Health Data clustering is a significant research direction aiming to extract knowledge from a continuous health data flow to support online health decisions. However, processing health data clusters is still a challenging task. Existing clustering approaches are subject to various limitations in terms of considering the neighbor clusters and conducting multiple operations during the maintenance process. In this paper, we model, design and implement a novel framework called IClustMaint for efficiently clustering and maintaining health data clusters incrementally. A two-phase algorithm is embedded in the framework. We first employ the Principal Component Analysis (PCA) method to efficiently reduce the high costs of the initial clustering phase. Next, in the maintenance phase, we propose the incremental Cluster maintenance (ICM) approach for managing the generated cluster during a period of time. Technically, when the data clusters are evolving over time and need to be maintained frequently, the ICM approach improves the performance of cluster maintenance by only tracking the edge points. The experimental results on a real medical dataset verify the efficiency of the proposed approaches.

Immune checkpoint blockade therapy mitigates systemic inflammation and affects cellular FLIP-expressing monocytic myeloid-derived suppressor cells in non-progressor non-small cell lung cancer patients

ABSTRACT Cancer cells favor the generation of myeloid cells with immunosuppressive and inflammatory features, including myeloid-derived suppressor cells (MDSCs), which support tumor progression. The anti-apoptotic molecule, cellular FLICE (FADD-like interleukin-1β-converting enzyme)-inhibitory protein (c-FLIP), which acts as an important modulator of caspase-8, is required for the development and function of monocytic (M)-MDSCs. Here, we assessed the effect of immune checkpoint inhibitor (ICI) therapy on systemic immunological landscape, including FLIP-expressing MDSCs, in non-small cell lung cancer (NSCLC) patients. Longitudinal changes in peripheral immunological parameters were correlated with patients’ outcome. In detail, 34 NSCLC patients were enrolled and classified as progressors (P) or non-progressors (NP), according to the RECIST evaluation. We demonstrated a reduction in pro-inflammatory cytokines such as IL-8, IL-6, and IL-1β in only NP patients after ICI treatment. Moreover, using t-distributed stochastic neighbor embedding (t-SNE) and cluster analysis, we characterized in NP patients a significant increase in the amount of lymphocytes and a slight contraction of myeloid cells such as neutrophils and monocytes. Despite this moderate ICI-associated alteration in myeloid cells, we identified a distinctive reduction of c-FLIP expression in M-MDSCs from NP patients concurrently with the first clinical evaluation (T1), even though NP and P patients showed the same level of expression at baseline (T0). In agreement with the c-FLIP expression, monocytes isolated from both P and NP patients displayed similar immunosuppressive functions at T0; however, this pro-tumor activity was negatively influenced at T1 in the NP patient cohort exclusively. Hence, ICI therapy can mitigate systemic inflammation and impair MDSC-dependent immunosuppression.