Unsupervised Clustering Techniques Research Articles

Enabling Reliable Visual Detection of Chronic Myocardial Infarction with Native T1 Cardiac MRI Using Data-Driven Native Contrast Mapping.

Purpose To investigate whether infarct-to-remote myocardial contrast can be optimized by replacing generic fitting algorithms used to obtain native T1 maps with a data-driven machine learning pixel-wise approach in chronic reperfused infarct in a canine model. Materials and Methods A controlled large animal model (24 canines, equal male and female animals) of chronic myocardial infarction with histologic evidence of heterogeneous infarct tissue composition was studied. Unsupervised clustering techniques using self-organizing maps and t-distributed stochastic neighbor embedding were used to analyze and visualize native T1-weighted pixel-intensity patterns. Deep neural network models were trained to map pixel-intensity patterns from native T1-weighted image series to corresponding pixels on late gadolinium enhancement (LGE) images, yielding visually enhanced noncontrast maps, a process referred to as data-driven native mapping (DNM). Pearson correlation coefficients and Bland-Altman analyses were used to compare findings from the DNM approach against standard T1 maps. Results Native T1-weighted images exhibited distinct pixel-intensity patterns between infarcted and remote territories. Granular pattern visualization revealed higher infarct-to-remote cluster separability with LGE labeling as compared with native T1 maps. Apparent contrast-to-noise ratio from DNM (mean, 15.01 ± 2.88 [SD]) was significantly different from native T1 maps (5.64 ± 1.58; P < .001) but similar to LGE contrast-to-noise ratio (15.51 ± 2.43; P = .40). Infarcted areas based on LGE were more strongly correlated with DNM compared with native T1 maps (R2 = 0.71 for native T1 maps vs LGE; R2 = 0.85 for DNM vs LGE; P < .001). Conclusion Native T1-weighted pixels carry information that can be extracted with the proposed DNM approach to maximize image contrast between infarct and remote territories for enhanced visualization of chronic infarct territories. Keywords: Chronic Myocardial Infarction, Cardiac MRI, Data-Driven Native Contrast Mapping Supplemental material is available for this article. © RSNA, 2024.

Bioinformatics Identification and Experimental Validation of a Prognostic Model for the Survival of Lung Squamous Cell Carcinoma Patients.

Lung squamous cell carcinoma (LUSC) kills more than four million people yearly. Creating more trustworthy tumor molecular markers for LUSC early detection, diagnosis, prognosis, and customized treatment is essential. Cuproptosis, a novel form of cell death, opened up a new field of study for searching for trustworthy tumor indicators. Our goal was to build a risk model to assess drug sensitivity, monitor immune function, and predict prognosis in LUSC patients. The 19 cuproptosis-related genes were found in the literature, and patient genomic and clinical information was collected using the Cancer Genomic Atlas (TCGA) database. The LUSC patients were grouped using unsupervised clustering techniques, and 7626 differentially expressed genes were identified. Using univariate COX analysis, LASSO regression analysis, and multivariate COX analysis, a prognostic model for LUSC patients was developed. The tumor immune escape was evaluated using the Tumor Immune Dysfunction and Exclusion (TIDE) method. The R packages 'pRRophetic,' 'ggpubr,' and 'ggplot2' were utilized to examine drug sensitivity. For modeling, a 6-cuproptosis-based gene signature was found. Patients with high-risk LUSC had significantly worse survival rates than those with low-risk conditions. The possibility of tumor immunological escape was increased in patients with higher risk scores due to more immune cell inactivation. For patients with high-risk LUSC, we discovered seven potent potential drugs (AZD6482, CHIR.99021, CMK, Embelin, FTI.277, Imatinib, and Pazopanib). In conclusion, the cuproptosis-based genes predictive risk model can be utilized to predict outcomes, track immune function, and evaluate medication sensitivity in LUSC patients.

Impact of COVID-19 pandemic on physician-scientist trainees to faculty one year into the pandemic

PurposePhysician-scientists play a crucial role in advancing biomedical sciences. Proportionally fewer physicians are actively engaged in scientific pursuits, attributed to attrition in the training and retention pipeline. This national study evaluated the ongoing and longer-term impact of the COVID-19 pandemic on stress levels, research productivity, and optimism for physician-scientists at all levels of training.MethodsA multi-institutional cross-sectional survey of medical students, graduate students, and residents/fellows/junior faculty (RFJF) was conducted from April to August 2021 to assess the impact of COVID-19 on individual stress, productivity, and optimism. Multivariate regression analyses were performed to identify associated variables and unsupervised variable clustering techniques were employed to identify highly correlated responses.ResultsA total 677 respondents completed the survey, representing different stages of physician-scientist training. Respondents report high levels of stress (medical students: 85%, graduate students: 63%, RFJF: 85%) attributed to impaired productivity concerns, concern about health of family and friends, impact on personal health and impairment in training or career development. Many cited impaired productivity (medical students: 65% graduate students: 79%, RFJF: 78%) associated with pandemic impacts on training, labs closures and loss of facility/resource access, and social isolation. Optimism levels were low (medical students: 37%, graduate students: 38% and RFJF: 39%) with females less likely to be optimistic and more likely to report concerns of long-term effects of COVID-19. Optimism about the future was correlated with not worrying about the long-term effects of COVID-19. Since the COVID-19 pandemic, all respondents reported increased prioritization of time with family/friends (67%) and personal health (62%) over career (25%) and research (24%).ConclusionsThis national survey highlights the significant and protracted impact of the COVID-19 pandemic on stress levels, productivity, and optimism among physician-scientists and trainees. These findings underscore the urgent need for tailored support, including mental health, academic, and career development assistance for this biomedical workforce.

Proposing unsupervised clustering-based earthquake records selection framework for computationally efficient nonlinear response history analysis of structures equipped with multi-stage friction pendulum bearings

The design and analysis of base-isolated structures are critical for ensuring seismic performance and safety. Nevertheless, to effectively design and analyze base-isolated structures, it is essential to consider the structure's response under various ground motions that may occur during an earthquake. The nonlinear response history analysis (NRHA) is a common method for evaluating structures' seismic response under a specific set of ground motions. However, conducting this analysis for many earthquakes can be computationally expensive and time-consuming. Therefore, it is desirable to have a method for selecting a subset of earthquake records representative of the earthquake population. Accordingly, this study introduces a novel block-based framework based on the K-means unsupervised clustering techniques for efficient earthquake records' selection NRHA of base-isolated structures. The proposed approach aims to reduce the number of records required for investigations by identifying clusters of similar seismic signals and selecting representative records from each cluster for the NRHA. Additionally, the proposed framework serves a vital role in identifying the earthquake records that can be used as a training dataset for developing artificial neural network (ANN) models, enabling rapid response estimation of base-isolated structures. This dual functionality of the proposed method provides significant value for researchers and engineers working in earthquake-prone countries. This article presents two case studies to demonstrate the applicability and effectiveness of the proposed framework. The first case study focuses on selecting a subset of earthquake records from a suit that meets the ASCE 7–22 requirements for NRHA on base-isolated structures, aiming to achieve the close mean responses between the full suit and the subset of records. The second case study highlights the development of an ANN model for response estimation of such structures using the proposed framework to select suitable earthquakes as the training dataset, hence providing an adequate database to achieve high estimation results. In general, the study results highlight the capabilities of the proposed frameworks and the benefits of the K-means unsupervised clustering approach in addressing the gap in the literature and providing a robust method for earthquake record selection in various analysis scenarios.

Optimal policies for nutrition administration to very low birth weight infants

AbstractVery low birth weight (VLBW) infants (birth weight 1500 grams) are at risk of postnatal growth restriction. Understanding how nutrition is associated with growth and how these associations vary based on infant characteristics and comorbidities is important to reduce postnatal growth restriction. We propose a three‐step analytical framework: (i) We use unsupervised Clustering techniques to identify subgroups within a cohort of VLBW infants based on infant characteristics, diagnoses, and treatments. (ii) For each cluster, we use Multilevel Modeling to explore the associations between calorie or protein intake and growth velocity (GV) for varying time windows. (iii) We build Mixed‐Integer Programming Models to achieve simple rule‐based policies that physicians can use to classify infants into one of the identified subgroups. We use electronic health records from VLBW infants at Lurie Children's Hospital in Chicago, IL, born between 2011 and 2014. We find that clustering separates infants into two clusters, with Cluster 1 having smaller infants with more comorbidities than Cluster 2. Initial clustering on only sex and birth weight provides results similar to clustering on later‐life diagnoses and treatments. Multilevel models with Clustering provide better model fit than models without clustering. For Cluster 1, there is a significant association between GV and protein but not calories. For Cluster 2, both protein and calories are individually associated with growth. We develop accurate and sparse scoring systems to help clinicians identify infants at higher risk of growth restriction and consider nutrition regimens accordingly.

Predictive and Prognostic Relevance of Tumor-Infiltrating Immune Cells: Tailoring Personalized Treatments against Different Cancer Types.

TIICs are critical components of the TME and are used to estimate prognostic and treatment responses in many malignancies. TIICs in the tumor microenvironment are assessed and quantified by categorizing immune cells into three subtypes: CD66b+ tumor-associated neutrophils (TANs), FoxP3+ regulatory T cells (Tregs), and CD163+ tumor-associated macrophages (TAMs). In addition, many cancers have tumor-infiltrating M1 and M2 macrophages, neutrophils (Neu), CD4+ T cells (T-helper), CD8+ T cells (T-cytotoxic), eosinophils, and mast cells. A variety of clinical treatments have linked tumor immune cell infiltration (ICI) to immunotherapy receptivity and prognosis. To improve the therapeutic effectiveness of immune-modulating drugs in a wider cancer patient population, immune cells and their interactions in the TME must be better understood. This study examines the clinicopathological effects of TIICs in overcoming tumor-mediated immunosuppression to boost antitumor immune responses and improve cancer prognosis. We successfully analyzed the predictive and prognostic usefulness of TIICs alongside TMB and ICI scores to identify cancer's varied immune landscapes. Traditionally, immune cell infiltration was quantified using flow cytometry, immunohistochemistry, gene set enrichment analysis (GSEA), CIBERSORT, ESTIMATE, and other platforms that use integrated immune gene sets from previously published studies. We have also thoroughly examined traditional limitations and newly created unsupervised clustering and deconvolution techniques (SpatialVizScore and ProTICS). These methods predict patient outcomes and treatment responses better. These models may also identify individuals who may benefit more from adjuvant or neoadjuvant treatment. Overall, we think that the significant contribution of TIICs in cancer will greatly benefit postoperative follow-up, therapy, interventions, and informed choices on customized cancer medicines.

Brain Tumor Detection Using Classification Model

This study aims to develop automated methods for detecting and segmenting brain tumors in MRI scans using neural networks and conventional image processing. Accurately identifying tumor location and boundaries is crucial for diagnosis and determining the most effective treatment approach. Manual review by radiologists is time-consuming and subjective. However, recent advances in deep learning now enable more efficient analysis of large MRI datasets.&#x0D; A variety of neural network models have been explored for tumor segmentation, including recurrent networks, autoencoders and convolutional networks. RNNs are well-suited to capture contextual information across sequential MRI slices. Autoencoders perform unsupervised feature learning from the images. CNNs have shown promise due to their ability to directly learn visual patterns from pixel values. Unsupervised clustering techniques such as fuzzy c-means and k-means are also investigated to group tumor pixels based on intensity similarity without labeled data.&#x0D; Previous research combining these techniques has achieved high segmentation accuracy between 84-97% using additional image processing steps. Automated segmentation provides benefits over manual inspection such as reduced workload and more consistent delineation of tumor margins. It can precisely outline growths to aid radiologists in diagnosis and pathologists in assessing malignancy.&#x0D; The detection and segmentation of brain tumors has applications in both radiology and oncology. Radiologists could benefit from faster scan review and automated report generation. Neurosurgeons would gain insights into a tumor's size, location and infiltration to determine the most suitable surgical approach. Oncologists could utilize quantitative tumor features to select the most targeted radiation therapy or chemotherapy protocols. Overall, continued advances in deep and machine learning hold promise to improve diagnosis and management of brain cancer patients.

Using students’ cognitive, affective, and demographic characteristics to predict their understanding of computational thinking concepts: a machine learning-based approach

ABSTRACT Logistic regression models have traditionally been used to identify the factors contributing to students’ conceptual understanding. With the advancement of the machine learning-based research approach, there are reports that some machine learning algorithms outperform logistic regression models in terms of prediction. In this study, we collected cognitive, affective, and demographic data from 4,142 primary students as features to predict their understanding of computational thinking (CT) concepts. We identified and used five popularly used machine learning models. All five machine learning models outperformed the logistic regression model, with the extreme gradient boosting (XGBoost) model achieving the highest predictive accuracy. We used these features and the K-means algorithm of the unsupervised clustering technique to identify four optimal clusters of students. By comparing the representative students from each of the four clusters, selected by the t-distributed stochastic neighbour embedding algorithm, we found that each cluster had its characteristics. For example, one cluster consisted of students who outperformed students in the other three clusters in mathematics, scored highest in prior experience in programming, and used computers and the Internet most frequently. By identifying the characteristics of these clusters, pedagogical design, and resource support can be proposed to support students’ learning of CT concepts.

AutoSCAN: automatic detection of DBSCAN parameters and efficient clustering of data in overlapping density regions.

The density-based clustering method is considered a robust approach in unsupervised clustering technique due to its ability to identify outliers, form clusters of irregular shapes and automatically determine the number of clusters. These unique properties helped its pioneering algorithm, the Density-based Spatial Clustering on Applications with Noise (DBSCAN), become applicable in datasets where various number of clusters of different shapes and sizes could be detected without much interference from the user. However, the original algorithm exhibits limitations, especially towards its sensitivity on its user input parameters minPts and ɛ. Additionally, the algorithm assigned inconsistent cluster labels to data objects found in overlapping density regions of separate clusters, hence lowering its accuracy. To alleviate these specific problems and increase the clustering accuracy, we propose two methods that use the statistical data from a given dataset's k-nearest neighbor density distribution in order to determine the optimal ɛ values. Our approach removes the burden on the users, and automatically detects the clusters of a given dataset. Furthermore, a method to identify the accurate border objects of separate clusters is proposed and implemented to solve the unpredictability of the original algorithm. Finally, in our experiments, we show that our efficient re-implementation of the original algorithm to automatically cluster datasets and improve the clustering quality of adjoining cluster members provides increase in clustering accuracy and faster running times when compared to earlier approaches.

Smart science: How artificial intelligence is revolutionizing pharmaceutical medicine

Abstract Artificial intelligence (AI) is a discipline within the field of computer science that encompasses the development and utilization of machines capable of emulating human behavior, particularly regarding the astute examination and interpretation of data. AI operates through the utilization of specialized algorithms, and it includes techniques such as deep (DL), and machine learning (ML), and natural language processing (NLP). As a result, AI has found its application in the study of pharmaceutical chemistry and healthcare. The AI models employed encompass a spectrum of methodologies, including unsupervised clustering techniques applied to drugs or patients to discern potential drug compounds or appropriate patient cohorts. Additionally, supervised ML methodologies are utilized to enhance the efficacy of therapeutic drug monitoring. Further, AI-aided prediction of the clinical outcomes of clinical trials can improve efficiency by prioritizing therapeutic intervention that are likely to succeed, hence benefiting the patient. AI may also help create personalized treatments by locating potential intervention targets and assessing their efficacy. Hence, this review provides insights into recent advances in the application of AI and different tools used in the field of pharmaceutical medicine.

Ensemble Clustering to Generate Phenotypes of Kidney Transplant Donors and Recipients.

In this paper we investigate the generation of phenotypes for kidney transplant donors and recipients to assist with decision making around organ allocation. We present an ensemble clustering approach for multi-type data (numerical and categorical) using two different clustering approaches-i.e., model based and vector quantization based clustering. These clustering approaches were applied to a large, US national deceased donor kidney transplant recipient database to characterize members of each cluster (in an unsupervised fashion) and to determine whether the subsequent risk of graft failure differed for each cluster. We generated three distinct clusters of recipients, which were subsequently used to generate phenotypes. Each cluster phenotype had recipients with varying clinical features, and the risk of kidney transplant graft failure and mortality differed across clusters. Importantly, the clustering results by both approaches demonstrated a significant overlap. Utilization of two distinct clustering approaches may be a novel way to validate unsupervised clustering techniques and clustering can be used for organ allocation decision making on the basis of differential outcomes.

Comparison of geological clusters between influenza and COVID-19 in Thailand with unsupervised clustering analysis.

The coronavirus disease (COVID-19) pandemic has considerably impacted public health, including the transmission patterns of other respiratory pathogens, such as the 2009 pandemic influenza (H1N1). COVID-19 and influenza are both respiratory infections that started with a lack of vaccination-based immunity in the population. However, vaccinations have been administered over time, resulting in a transition of the status of both diseases from a pandemic to an endemic. In this study, unsupervised clustering techniques were used to identify clusters of disease trends in Thailand. The analysis incorporated three distinct surveillance datasets: the pandemic influenza outbreak, influenza in the endemic stage, and the early stages of COVID-19. The analysis demonstrated a significant difference in the distribution of provinces between Cluster -1, representing those with unique transmission patterns, and the other clusters, indicating provinces with similar transmission patterns among their members. Specifically, for Pandemic Influenza, the ratio was 61:16, while for Pandemic COVID-19, it was 65:12. In contrast, Endemic Influenza exhibited a ratio of 46:31, with a notable emergence of more clustered provinces in the southern, western, and central regions. Furthermore, a pair of provinces with highly similar spreading patterns were identified during the pandemic stages of both influenza and COVID-19. Although the similarity decreased slightly for endemic influenza, they still belonged to the same cluster. Our objective was to identify the transmission patterns of influenza and COVID-19, with the aim of providing quantitative and spatial information to aid public health management in preparing for future pandemics or transitioning into an endemic phase.

Diagnostic utility and reporting recommendations for clinical DNA methylation episignature testing in genetically undiagnosed rare diseases

PurposeThis study aims to assess the diagnostic utility and provide reporting recommendations for clinical DNA methylation episignature testing based on the cohort of patients tested through the EpiSign Clinical Testing Network. MethodsThe EpiSign assay utilized unsupervised clustering techniques and a support vector machine–based classification algorithm to compare each patient’s genome-wide DNA methylation profile with the EpiSign Knowledge Database, yielding the result that was reported. An international working group, representing distinct EpiSign Clinical Testing Network health jurisdictions, collaborated to establish recommendations for interpretation and reporting of episignature testing. ResultsAmong 2399 cases analyzed, 1667 cases underwent a comprehensive screen of validated episignatures, imprinting, and promoter regions, resulting in 18.7% (312/1667) positive reports. The remaining 732 referrals underwent targeted episignature analysis for assessment of sequence or copy-number variants (CNVs) of uncertain significance or for assessment of clinical diagnoses without confirmed molecular findings, and 32.4% (237/732) were positive. Cases with detailed clinical information were highlighted to describe various utility scenarios for episignature testing. ConclusionClinical DNA methylation testing including episignatures, imprinting, and promoter analysis provided by an integrated network of clinical laboratories enables test standardization and demonstrates significant diagnostic yield and clinical utility beyond DNA sequence analysis in rare diseases.

Advancing Fingerprint Template Generation and Matching With Recast Minutiae Clustering and mRBFN

Rapid development of automation in the day-to-day life activity marks up the need of securing bio-metric template and the privacy of rightful owner. Minutiae-based matching is the most popular in the fingerprint recognition system, which greatly suffers from non-linear distortion like translation and rotation. To deal with linear distortion most of the technique proposed in the literature depends upon a reference or singular point. The paper proposes a binary template generation technique which applies an unsupervised clustering technique without fixing the number of clusters. Instead of position and orientation of the minutiae points the cardinality of the clusters are stored and converted into binary template. No spatial pattern information about the fingerprint is stored in the template to protect it from spoofing and information leakage. By the help of modified Radial Basis Function Network (mRBFN) with robust and efficient matching technique the generated templates are matched for authentication. We use MCYT dataset for training the mRBFN. The efficiency of the proposed scheme is evaluated on FVC 2000, FVC 2002 and FVC 2004 dataset.

Doğrudan Yabancı Yatırım-Gelir Eşitsizliği Bağlamı ve Özümseme Kapasitesinin Rolü: Bir Sonlu Karışım Model Yaklaşımı

The income inequality-FDI nexus is ambiguous on theoretical grounds, as several opposing mechanisms relating FDI to inequality are put forward in the theoretical literature. Empirical studies also produce mixed results. It might suggest a heterogeneous response of income inequality to FDI inflows conditional on distinct characteristics in recipient countries. Although unobserved country-specific characteristics are considered by fixed or random effects modeling in panel regression applications, some studies incorporate observable country-specific factors such as absorptive capacity as a conditioning factor to explain the reasons for conflicting findings. Therefore, the existing studies employ threshold panel regression models that split the sample based on the outcome (supervised learning). Differently from the previous ones, this study takes a distinct empirical strategy by adopting a finite mixture model (FMM) as an unsupervised model-based clustering technique to scrutinize distributional heterogeneity in the linkage between FDI and income inequality. The study then questions the role of absorptive capacity as a conditioning factor with varying effects on the inequality of FDI. Our empirical results, based on panel data from 26 developing countries between 2004-2019, explain the varying effects of FDI on income inequality according to three different country clusters. FDI improves income inequality in the first cluster, while it does not significantly affect income inequality in the second and deteriorates income inequality in the third cluster. Furthermore, our empirical findings reveal that a country's high absorptive capacity, especially its high level of human capital, prevents the negative impact of FDI on income distribution.

Dynamic Population on Bio-Inspired Algorithms Using Machine Learning for Global Optimization.

In the optimization field, the ability to efficiently tackle complex and high-dimensional problems remains a persistent challenge. Metaheuristic algorithms, with a particular emphasis on their autonomous variants, are emerging as promising tools to overcome this challenge. The term "autonomous" refers to these variants' ability to dynamically adjust certain parameters based on their own outcomes, without external intervention. The objective is to leverage the advantages and characteristics of an unsupervised machine learning clustering technique to configure the population parameter with autonomous behavior, and emphasize how we incorporate the characteristics of search space clustering to enhance the intensification and diversification of the metaheuristic. This allows dynamic adjustments based on its own outcomes, whether by increasing or decreasing the population in response to the need for diversification or intensification of solutions. In this manner, it aims to imbue the metaheuristic with features for a broader search of solutions that can yield superior results. This study provides an in-depth examination of autonomous metaheuristic algorithms, including Autonomous Particle Swarm Optimization, Autonomous Cuckoo Search Algorithm, and Autonomous Bat Algorithm. We submit these algorithms to a thorough evaluation against their original counterparts using high-density functions from the well-known CEC LSGO benchmark suite. Quantitative results revealed performance enhancements in the autonomous versions, with Autonomous Particle Swarm Optimization consistently outperforming its peers in achieving optimal minimum values. Autonomous Cuckoo Search Algorithm and Autonomous Bat Algorithm also demonstrated noteworthy advancements over their traditional counterparts. A salient feature of these algorithms is the continuous nature of their population, which significantly bolsters their capability to navigate complex and high-dimensional search spaces. However, like all methodologies, there were challenges in ensuring consistent performance across all test scenarios. The intrinsic adaptability and autonomous decision making embedded within these algorithms herald a new era of optimization tools suited for complex real-world challenges. In sum, this research accentuates the potential of autonomous metaheuristics in the optimization arena, laying the groundwork for their expanded application across diverse challenges and domains. We recommend further explorations and adaptations of these autonomous algorithms to fully harness their potential.

Deep Proteomic Analysis Reveals Shared Terminal Mechanisms for Familial Hemophagocytic Lymphohistiocytosis and Lymphoma-Associated Hyperinflammation

Background: Hemophagocytic lymphohistiocytosis (HLH) is a hyperinflammatory syndrome increasingly observed in patients with hematologic malignancies (HM-HLH). Familial HLH (FHL) is seen in young children and is typically caused by mutations affecting the perforin pathway of lymphocyte cytotoxic function, leading to excessive CD8+ T cell activation and interferon-gamma production. Though HM-HLH has clinical similarities to FHL, little is known regarding the pathophysiology of HM-HLH. We previously developed the ‘Optimized HLH Inflammatory (OHI) index,‘ combining soluble CD25 and ferritin elevations, to improve recognition of hyperinflammation and prognostication in HM patients (Zoref-Lorenz et al., Blood; 2022). Based on these findings, we hypothesize that OHI+ HM's share similar immune pathophysiology with FHL and aim to validate this hypothesis using an unbiased proteomic approach. Methods: We conducted an unbiased proteomic analysis of samples from Schneider Children's, Meir, Sheba, and Beilinson Medical Centers. Our cohort included FHL patients, healthy pediatric controls, and patients with HMs, classified as HM-OHI+(sCD25&amp;gt;3,900 U/mL and ferritin&amp;gt;1,000 ng/mL) or HM-OHI-. SomaLogic proteomics platform, an aptamer-based method, assessed 7,596 serum proteins representing 6,412 genes. Unsupervised clustering techniques (primary component analysis and unbiased hierarchical clustering) were used to group patient results. Gene set enrichment analysis (GSEA) of the hallmark set was used to identify enriched biologic pathways in HM-OHI+ and FHL. Differential expression analysis (Student t-test) identified differentially expressed proteins (DEPs, q&amp;lt;0.1, p&amp;lt;0.05) between FHL vs. healthy controls and HM-OHI+ vs. HM-OHI-. Venn analysis revealed overlapping DEPs between FHL and OHI+ HM-HLH patients. GSEA with hallmark pathways evaluated the overlapping genes in each pairwise comparison within the entire sets and the intersecting genes. Spearman's correlation analysis examined the relationship between fold changes of proteins and normalized enrichment scores of hallmark GSEA comparisons within intersecting proteins and genes of FHL and HM-OHI+. Logistic regression created an equation for the linear curve of these correlations. Results: We analyzed 42 patients, including 14 HM-OHI+ and 14 HM-OHI- patients, matched by malignancy type, 7 FHL patients, and age-matched controls. Unsupervised hierarchical clustering (Figure 1A) showed that HM-OHI+ patients clustered with FHL patients, while healthy controls and HM-OHI- patients clustered together. GSEA analysis identified common enriched biological pathways between HM-OHI+ and FHL: interferon-γ response, IL6 JAK STAT3 signaling, IL2 STAT5 signaling, and oxidative phosphorylation (Normalized enrichment scores of 1.82:1.92; 1.76:2.23; 1.76:1.76 and 1.68:1.78 respectively for HM-OHI+ and FHL). In pairwise statistical comparisons, 841 differential proteins were shared between FHL and HM-OHI+. HM-OHI+ and FHL exhibited highly correlated fold changes (R=0.97, p&amp;lt;0.0001, Y= 1.362 X-0.2597, Figure 1B), with highly significant elevated proteins in HM-OHI+ vs. HM-OHI- that are key in the mechanism of FHL: T cell activation (GRZB, FC-1.6 q=0.01; GRZH: FC-1.5, q= 0.002), IFN-γ response (CXCL10, FC-2.1,q=0.001), and macrophage activation (CD14: FC- 1.6, q=0.001; CSF1:FC- 2.1, 1&amp;lt;0.0001; IL1RL: FC- 1.8,q=0.0009). The intersecting genes in the GSEA analysis revealed a high correlation of the biological pathways (R=0.94, p&amp;lt;0.0001, Y=1.282X+0), with the IFN-γ response, the known driver of FHL being the most significantly over-enriched pathway in both pairwise comparisons (q&amp;lt;0.25). These findings indicate shared pathophysiology mechanisms between HM-OHI+ and FHL. We validated key proteins using ELISA as an orthogonal method. Conclusions: This study establishes a common downstream mechanism of hyperinflammation in hematologic malignancies and FHL, closely linked to the IFN-γ response, and validates the OHI index as a diagnostic tool for HM-HLH. Understanding the shared pathophysiology between HM-OHI+ and FHL sheds light on the poor prognosis of HM-associated hyperinflammation and emphasizes the need to study the upstream mechanisms of HM-HLH to define targets for improved therapeutic approaches.

Low-Cost Optimized U-Net Model with GMM Automatic Labeling Used in Forest Semantic Segmentation.

Currently, Convolutional Neural Networks (CNN) are widely used for processing and analyzing image or video data, and an essential part of state-of-the-art studies rely on training different CNN architectures. They have broad applications, such as image classification, semantic segmentation, or face recognition. Regardless of the application, one of the important factors influencing network performance is the use of a reliable, well-labeled dataset in the training stage. Most of the time, especially if we talk about semantic classification, labeling is time and resource-consuming and must be done manually by a human operator. This article proposes an automatic label generation method based on the Gaussian mixture model (GMM) unsupervised clustering technique. The other main contribution of this paper is the optimization of the hyperparameters of the traditional U-Net model to achieve a balance between high performance and the least complex structure for implementing a low-cost system. The results showed that the proposed method decreased the resources needed, computation time, and model complexity while maintaining accuracy. Our methods have been tested in a deforestation monitoring application by successfully identifying forests in aerial imagery.

Changepoint detection-assisted nonparametric clustering for unsupervised temporal sign segmentation

Temporal sign segmentation aims to temporarily divide continuous sign language into category-agnostic sign segments. One of the challenges to temporal sign segmentation is the paucity of frame-level annotations for sign language videos, which restricts the applicability of supervised and semi-supervised approaches. To address this challenge, we consider temporal sign segmentation as a clustering problem defined as the grouping of semantically similar frames of a given video, which allows us to choose unsupervised clustering techniques as promising alternatives. However, most unsupervised clustering techniques are parametric. In the context of temporal sign segmentation, they assume that the number of sign segments of a given continuous sign video is predefined, which is implausible. The primary contributions of this study are as follows: (1) We propose the first nonparametric clustering algorithm for unsupervised temporal sign segmentation. The main concept is to enhance the hierarchical graph-based clustering algorithm to be nonparametric by adopting cost and penalty functions of a changepoint detection algorithm to determine the optimal number of sign segments. Experimental results show that the performance of the proposed unsupervised method is comparable to that of the latest semi-supervised sign segmentation method in terms of several metrics. Moreover, the execution time of the proposed clustering method was less than 1 s, thereby ensuring its applicability. (2) We identify that the conventional metrics for temporal sign segmentation do not sufficiently address over-segmentation. To overcome the difficulty, we propose a new main metric to evaluate the performance of temporal sign segmentation, called the adjusted MF1B.

Natural language processing (NLP) aided qualitative method in health research

Qualitative data analysis is produced frequently in healthcare settings, which is a time-consuming and skilled analytic task. The use of qualitative research findings in clinical settings takes years, which is sometimes obsolete knowledge as the health context is dynamic. Artificial Intelligence (AI)-based qualitative data analysis might present with rapid analysis of text-based data in real-time, thereby empowering qualitative researchers to expedite their analysis and facilitate timely use of the research findings. We tested an AI-based method to complement the manual analysis of text-based data from the verbatim transcripts of seven mall managers’ interviews. First, we prepared text data into a machine-calculable format and employed BERT model to extract sentence-level features in our case. Second, we implement TF-IDF-based keywords mining techniques to extract the main candidate themes from the interview transcripts to support text-based analysis, including: 1) primary cluster detection algorithm, and 2) keyword extraction algorithm. The extracted core themes provide qualitative researchers with a more comprehensive overview of the qualitative data. Most of the sentences clustered in meaningful short topics or sentences carrying independent and clear information. The extracted topics and clustered sentences reduced qualitative researchers’ workload by condensing and identifying meaningful concepts and naming them. This method combining contextualized word embeddings, unsupervised clustering, and keyword extraction techniques can significantly reduce the overall workload and time consumed in qualitative research using conventional methods.