Unsupervised K-means Clustering Research Articles

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.

Data driven assessment of rock mass quality in red-bed hilly area: a case study of Guang’an city, SW China

The evaluation of geological suitability for urban underground space (UUS) development is an indispensable prerequisite for its optimal utilization. As the actual carrier of underground facilities, the evaluation of rock mass quality plays a crucial role in assessing geological suitability. However, it is notable that the evaluation of rock mass quality has regrettably remained somewhat marginalized within the broader framework of the geological suitability assessment in recent years. The selection of pertinent indicators for the evaluation of rock mass quality inherently presents an appreciable degree of subjectivity. Predominantly subjective evaluation methods continue to dominate the field, while the application of objective algorithms, such as unsupervised clustering, remains in its nascent stage. Furthermore, there is a lack of comprehensive investigations into distinct combinations of attributes. This limitation confines the broader applicability of the evaluation outcomes in the context of urban underground space. Within this study, we meticulously amassed rock core test data from over 40 boreholes of engineering geological significance within the urban planning ambit of Guang'An City. Utilizing the K-means unsupervised clustering algorithm and the Principal Component Analysis (PCA) algorithm. We successfully conducted an unsupervised clustering procedure with nine distinct physical and mechanical attributes. This yielded an aggregation into five discernible clusters. Building upon the derived clustering outcomes, a stratification of rock mass quality was effectuated into three distinct tiers: Level 1 (characterized by pure sandstone), Level 2 (primarily dominated by sandstone), and Level 3 (denoting fair conditions predominantly influenced by mudstone). This structured stratification facilitates a relatively objective and comprehensive evaluation of rock mass quality within the context of the red-bed hilly terrain. In the course of this analytical trajectory, we conducted a dissection of the clustering efficacy. For strongly correlated attributes, we propose a preliminary dimensionality reduction procedure prior to the clustering endeavor. Moreover, we recommend intervals of 10 m for the stratified evaluation in red bed hilly urban terrains.

Construction of a gene signature associated with anoikis to evaluate the prognosis and immune infiltration in patients with colorectal cancer.

Colorectal cancer (CRC) is characterized by a high metastasis rate, leading to poor prognosis and increased mortality. Anoikis, a physiological process, serves as a crucial barrier against metastasis. The objective of this research is to construct a prognostic model for CRC based on genes associated with anoikis. The study involved differential analysis and univariate Cox analysis of anoikis-related genes (ARGs), resulting in the selection of 47 genes closely associated with prognosis. Subsequently, unsupervised k-means clustering analysis was conducted on all patients to identify distinct clusters. Survival analysis, principal component analysis (PCA), and t-distributed stochastic neighbor embedding (t-SNE) analysis were performed on the different clusters to investigate associations within the clusters. Gene set variation analysis (GSVA) and gene set enrichment analysis (GSEA) were utilized to assess metabolic pathway enrichment between the identified clusters. Furthermore, single-sample GSEA (ssGSEA) was applied to explore variations in immune infiltration. Multivariable Cox regression and least absolute shrinkage and selection operator (LASSO) analyses were conducted to construct a risk model based on ten signatures, which enabled the grouping of all samples according to their risk scores. The prognostic value of the model was validated using receiver operating characteristic (ROC) curves, area under the curve (AUC) calculations, and survival curves. Additionally, the expression of candidate genes was validated using quantitative real-time polymerase chain reaction (qRT-PCR). Forty-seven survival-related ARGs were screened out. Somatic mutation analysis showed that these genes revealed a high mutation rate. Based on their expression, two clusters were identified. Cluster B patients exhibited a shortened overall survival and higher immune infiltration. A risk scoring model including ten genes was subsequently developed, which exhibited excellent prognostic predictive ability for CRC, as evidenced by the survival curve, ROC curve, and AUC curve. In addition, a nomogram was developed for predicting 3- and 5-year survival probabilities. The qRT-PCR results indicated the dissimilarities among the ten signatures in the tumor tissues and adjacent tissues of patients with CRC were fundamentally consistent with the analytical findings. This study comprehensively evaluated the prognostic significance of ARGs in CRC. It identified two distinct anoikis-related clusters and examined their respective immune microenvironments. Furthermore, an ARGs signature was developed to effectively predict the prognosis of CRC, thereby establishing a solid foundation for investigating the clinical prognostic role of anoikis in CRC.

Efficient Training of Colorectal Cancer Diagnosis Model through Unsupervised Learning Composite Network

The use of machine learning algorithms for precise and effective colorectal cancer diagnosis has gained popularity as a result of advancements in medical imaging. Through the use of an Unsupervised Learning Composite Network (ULCN) model this study suggests a novel method for training a diagnosis model. Conventional training techniques for these models frequently rely on sizable labelled datasets, which can be labour and resource-intensive to create. The ULCN, on the other hand, incorporates unsupervised learning into the training process, decreasing the need on labelled input. In order to address this issue, our study offers a revolutionary method known as the DL-Kmeans (Deep Learning with Unsupervised K-Means Clustering), which combines the effectiveness of the unsupervised K-means clustering algorithm with the power of deep learning. When compared to manual screening and annotation techniques, the DL-Kmeans ability to more quickly refine medical images serves as evidence of its effectiveness. The use of DL- Kmeans processed photos resulted in a two-fold acceleration in the training process for deep learning models used to diagnose colorectal cancer, which is very notable. Notably, this speeding up did not degrade performance quality; the DL- Kmeans enhanced method showed superior results in terms of training loss and accuracy attained. DL- Kmeans importance goes beyond only improving images. It is a useful tool for handling the growing volume of medical photos, which will help with the later creation of artificial intelligence models.

Application of machine learning for identification of heterotic groups in sunflower through combined approach of phenotyping, genotyping and protein profiling

Application of machine learning in plant breeding is a recent concept, that has to be optimized for precise utilization in the breeding program of high yielding crop plants. Identification and efficient utilization of heterotic grouping pattern aided with machine learning approaches is of utmost importance in hybrid cultivar breeding as it can save time and resources required to breed a new plant hybrid/variety. In the present study, 109 genotypes of sunflower were investigated at morphological, biochemical (SDS-PAGE) and molecular levels (through micro-satellites (SSR) markers) for heterotic grouping. All the three datasets were combined, scaled, and subjected to unsupervised machine learning algorithms, i.e., Hierarchical clustering, K-means clustering and hybrid clustering algorithm (hierarchical + K-means) for assessment of efficiency and resolution power of these algorithms in practical plant breeding for heterotic grouping identification. Following the application of machine learning unsupervised clustering approach, two major groups were identified in the studied sunflower germplasm, and further classification revealed six smaller classes in each major group through hierarchical and hybrid clustering approach. Due to high resolution, obtained in hierarchical clustering, classification achieved through this algorithm was further used for selection of potential parents. One genotype from each smaller group was selected based on the maximum seed yield potential and hybridized in a line × tester mating design producing 36 F1 cross combinations. These F1s along with their parents were studied in open field conditions for validating the efficacy of identified heterotic groups in sunflowers genetic material under study. Data for 11 agronomic and qualitative traits were recorded. These 36 F1 combinations were tested for their combining ability (General/Specific), heterosis, genotypic and phenotypic correlation and path analysis. Results suggested that F1 hybrids performed better for all the traits under investigation than their respective parents. Findings of the study validated the use of machine learning approaches in practical plant breeding; however, more accurate and robust clustering algorithms need to be developed to handle the data noisiness of open field experiments.

Spatial transcriptomics reveals alterations in perivascular macrophage lipid metabolism in the onset of Wooden Breast myopathy in broiler chickens

This study aims to use spatial transcriptomics to characterize the cell-type-specific expression profile associated with the microscopic features observed in Wooden Breast myopathy. 1 cm3 muscle sample was dissected from the cranial part of the right pectoralis major muscle from three randomly sampled broiler chickens at 23 days post-hatch and processed with Visium Spatial Gene Expression kits (10X Genomics), followed by high-resolution imaging and sequencing on the Illumina Nextseq 2000 system. WB classification was based on histopathologic features identified. Sequence reads were aligned to the chicken reference genome (Galgal6) and mapped to histological images. Unsupervised K-means clustering and Seurat integrative analysis differentiated histologic features and their specific gene expression pattern, including lipid laden macrophages (LLM), unaffected myofibers, myositis and vasculature. In particular, LLM exhibited reprogramming of lipid metabolism with up-regulated lipid transporters and genes in peroxisome proliferator-activated receptors pathway, possibly through P. Moreover, overexpression of fatty acid binding protein 5 could enhance fatty acid uptake in adjacent veins. In myositis regions, increased expression of cathepsins may play a role in muscle homeostasis and repair by mediating lysosomal activity and apoptosis. A better knowledge of different cell-type interactions at early stages of WB is essential in developing a comprehensive understanding.

Validation of Distinct Bladder Pain Phenotypes Utilizing the MAPP Research Network Cohort

Introduction and HypothesisAs interstitial cystitis/bladder pain syndrome (IC/BPS) likely represents multiple pathophysiologies, we sought to validate three clinical phenotypes of IC/BPS patients in a large, multi-center cohort using unsupervised machine learning (ML) analysis.MethodsUsing the female Genitourinary Pain Index and O’Leary-Sant Indices, k-means unsupervised clustering was utilized to define symptomatic phenotypes in 130 premenopausal IC/BPS participants recruited through the Multidisciplinary Approach to the Study of Chronic Pelvic Pain (MAPP) research network. Patient-reported symptoms were directly compared between MAPP ML-derived phenotypic clusters to previously defined phenotypes from a single center (SC) cohort.ResultsUnsupervised ML categorized IC/BPS participants into three phenotypes with distinct pain and urinary symptom patterns: myofascial pain, non-urologic pelvic pain, and bladder-specific pain. Defining characteristics included presence of myofascial pain or trigger points on examination for myofascial pain patients (p = 0.003) and bladder pain/burning for bladder-specific pain patients (p < 0.001). The three phenotypes were derived using only 11 features (fGUPI subscales and ICSI/ICPI items), in contrast to 49 items required previously. Despite substantial reduction in classification features, unsupervised ML independently generated similar symptomatic clusters in the MAPP cohort with equivalent symptomatic patterns and physical examination findings as the SC cohort.ConclusionsThe reproducible identification of IC/BPS phenotypes, distinguishing bladder-specific pain from myofascial and genital pain, using independent ML analysis of a multicenter database suggests these phenotypes reflect true pathophysiologic differences in IC/BPS patients.

The immune landscape of sepsis and using immune clusters for identifying sepsis endotypes.

The dysregulated immune response to sepsis still remains unclear. Stratification of sepsis patients into endotypes based on immune indicators is important for the future development of personalized therapies. We aimed to evaluate the immune landscape of sepsis and the use of immune clusters for identifying sepsis endotypes. The indicators involved in innate, cellular, and humoral immune cells, inhibitory immune cells, and cytokines were simultaneously assessed in 90 sepsis patients and 40 healthy controls. Unsupervised k-means cluster analysis of immune indicator data were used to identify patient clusters, and a random forest approach was used to build a prediction model for classifying sepsis endotypes. We depicted that the impairment of innate and adaptive immunity accompanying increased inflammation was the most prominent feature in patients with sepsis. However, using immune indicators for distinguishing sepsis from bacteremia was difficult, most likely due to the considerable heterogeneity in sepsis patients. Cluster analysis of sepsis patients identified three immune clusters with different survival rates. Cluster 1 (36.7%) could be distinguished from the other clusters as being an "effector-type" cluster, whereas cluster 2 (34.4%) was a "potential-type" cluster, and cluster 3 (28.9%) was a "dysregulation-type" cluster, which showed the lowest survival rate. In addition, we established a prediction model based on immune indicator data, which accurately classified sepsis patients into three immune endotypes. We depicted the immune landscape of patients with sepsis and identified three distinct immune endotypes with different survival rates. Cluster membership could be predicted with a model based on immune data.

Objective characterisation of reinforced concrete with progressive corrosion defects through clustering and thresholding of infrared images

The research in this paper focuses on detection and quantification of subsurface damage in reinforced concrete (RC) structures by the analysis of infrared images. Experimental investigations were performed on RC slabs with embedded reinforcing bars (rebars). The corrosion mechanism in rebars was accelerated using an electrical circuit setup. The initial temperature of the slabs was adjusted in the environmental chamber and IR images were taken at regular intervals while they were exchanging heat to laboratory air through convection. Records of IR images were post-processed using an objective thresholding method proposed based on unsupervised k-means clustering. The results were discussed in context and with respect to the ASTM D 4788–03 recommendation for minimum detectable thermal contrast (0.5 °C) from delamination in bridge concrete by infrared thermography (IRT). The conclusions lead to recommendations for improved practical implementation of IRT that contribute to the broader prospects of structural health monitoring (SHM).

Towards user-centric BCI design: Markov chain-based user assessment for mental imagery EEG-BCIs

Objective. While electroencephalography (EEG)-based brain–computer interfaces (BCIs) have many potential clinical applications, their use is impeded by poor performance for many users. To improve BCI performance, either via enhanced signal processing or user training, it is critical to understand and describe each user’s ability to perform mental control tasks and produce discernible EEG patterns. While classification accuracy has predominantly been used to assess user performance, limitations and criticisms of this approach have emerged, thus prompting the need to develop novel user assessment approaches with greater descriptive capability. Here, we propose a combination of unsupervised clustering and Markov chain models to assess and describe user skill. Approach. Using unsupervised K-means clustering, we segmented the EEG signal space into regions representing pattern states that users could produce. A user’s movement through these pattern states while performing different tasks was modeled using Markov chains. Finally, using the steady-state distributions and entropy rates of the Markov chains, we proposed two metrics taskDistinct and relativeTaskInconsistency to assess, respectively, a user’s ability to (i) produce distinct task-specific patterns for each mental task and (ii) maintain consistent patterns during individual tasks. Main results. Analysis of data from 14 adolescents using a three-class BCI revealed significant correlations between the taskDistinct and relativeTaskInconsistency metrics and classification F1 score. Moreover, analysis of the pattern states and Markov chain models yielded descriptive information regarding user performance not immediately apparent from classification accuracy. Significance. Our proposed user assessment method can be used in concert with classifier-based analysis to further understand the extent to which users produce task-specific, time-evolving EEG patterns. In turn, this information could be used to enhance user training or classifier design.

Computed tomography imaging phenotypes of hepatoblastoma identified from radiomics signatures are associated with the efficacy of neoadjuvant chemotherapy

BackgroundThough neoadjuvant chemotherapy has been widely used in the treatment of hepatoblastoma, there still lacks an effective way to predict its effect.ObjectiveTo characterize hepatoblastoma based on radiomics image features and identify radiomics-based lesion phenotypes by unsupervised machine learning, intended to build a classifier to predict the response to neoadjuvant chemotherapy.Materials and methodsIn this retrospective study, we segmented the arterial phase images of 137 cases of pediatric hepatoblastoma and extracted the radiomics features using PyRadiomics. Then unsupervised k-means clustering was applied to cluster the tumors, whose result was verified by t-distributed stochastic neighbor embedding (t-SNE). The least absolute shrinkage and selection operator (LASSO) regression was used for feature selection, and the clusters were visually analyzed by radiologists. The correlations between the clusters, clinical and pathological parameters, and qualitative radiological features were analyzed.ResultsHepatoblastoma was clustered into three phenotypes (homogenous type, heterogenous type, and nodulated type) based on radiomics features. The clustering results had a high correlation with response to neoadjuvant chemotherapy (P=0.02). The epithelial ratio and cystic components in radiological features were also associated with the clusters (P=0.029 and 0.008, respectively).ConclusionsThis radiomics-based cluster system may have the potential to facilitate the precise treatment of hepatoblastoma. In addition, this study further demonstrated the feasibility of using unsupervised machine learning in a disease without a proper imaging classification system.Graphical abstract

Image Segmentation and Filtering of Anaerobic Lagoon Floating Cover in Digital Elevation Model and Orthomosaics Using Unsupervised k-Means Clustering for Scum Association Analysis

In various engineering applications, remote sensing images such as digital elevation models (DEMs) and orthomosaics provide a convenient means of generating 3D representations of physical assets, enabling the discovery of new insights and analyses. However, the presence of noise and artefacts, particularly unwanted natural features, poses significant challenges, and their removal requires the application of filtering techniques prior to conducting analysis. Unmanned aerial vehicle-based photogrammetry is used at Melbourne Water’s Western Treatment Plant as a cost-effective and efficient method of inspecting the floating covers on the anaerobic lagoons. The focus of interest is the elevation profile of the floating covers for these sewage-processing lagoons and its implications for sub-surface scum accumulation, which can compromise the structural integrity of the engineered assets. However, unwanted artefacts due to trapped rainwater, debris, dirt, and other irrelevant structures can significantly distort the elevation profile. In this study, a machine learning algorithm is utilised to group distinct features on the floating cover based on an image segmentation process. An unsupervised k-means clustering algorithm is employed, which operates on a stacked 4D array composed of the elevation of the DEM and the RGB channels of the associated orthomosaic. In the cluster validation process, seven cluster groups were considered optimal based on the Calinski–Harabasz criterion. Furthermore, by utilising the k-means method as a filtering technique, three clusters contain features related to the elevations associated with the floating cover membrane, collectively representing 84% of the asset, with each cluster contributing at least 19% of the asset. The artefact groups constitute less than 6% of the asset and exhibit significantly different features, colour characteristics, and statistical measurements from those of the membrane groups. The study found notable improvements using the k-means filtering method, including a 59.4% average reduction in outliers and a 36.3% decrease in standard deviation compared to raw data. Additionally, employing the proposed method in the scum hardness analysis improved correlation strength by 13.1%, removing approximately 16% of the artefacts in total assets, in contrast to a 3.6% improvement with the median filtering method. This improved imaging will lead to significant benefits when integrating imagery into deep learning models for structural health monitoring and asset performance.

OR20-01 Machine Learning-based Steroid Metabolome Analysis In Women With Polycystic Ovary Syndrome Reveals Three Distinct Androgen Excess Subtypes With Different Metabolic Risk Profiles.

Abstract Disclosure: T.P. Rocha: None. E. Melson: None. R.J. Veen: None. L. Abdi: None. T. McDonnell: None. V. Tandl: None. J. Hawley: None. L. Wittemans: None. A. Anthony: None. L. Gilligan: None. F. Shaheen: None. P. Kempegowda: None. C.D. Gillett: None. L. Cussen: None. C. Missbrenner: None. F. Lajeunesse-Trempe: None. H. Gleeson: None. A. Rees: None. L. Robinson: None. C. Jayasenna: None. H.S. Randeva: None. H.S. Randeva: None. G.K. Dimitriadis: None. L.G. Gomes: None. A. Sitch: None. E. Vradi: Employee; Self; Bayer Schering Pharma. A.E. Taylor: None. M.W. O'Reilly: None. B.M. Obermayer-Pietsch: None. M. Biehl: None. W. Arlt: None. Introduction: Polycystic ovary syndrome affects 10% of women and is associated with an increased risk of type 2 diabetes, hypertension, and fatty liver disease. Androgen excess, a defining feature of PCOS, has been implicated as a driver of metabolic risk. Adrenal-derived 11-oxygenated androgens are a component of PCOS-related androgen excess and are preferentially activated in adipose tissue. Here, we aimed to identify PCOS sub-types with distinct steroid metabolomes and compare their cardiometabolic risk parameters. Methods: We prospectively recruited 488 treatment-naïve women with PCOS fulfilling diagnostic Rotterdam criteria [median age 28 (IQR 24-32) years; BMI 27.5 (22.4-34.6) kg/m2] at eight centres in the UK &amp; Ireland (n=208), Austria (n=242), and Brazil (n=38). Participants underwent a standardised assessment, including insulin sensitivity at baseline (HOMA-IR) and across a 2-h oGTT (Matsuda ISI). We used tandem mass spectrometry to measure 11 androgenic serum steroids, including classic and 11-oxygenated androgens. Results were analysed by unsupervised k-means clustering, followed by a comparison of clinical and metabolic phenotype parameters. Results: Machine learning analysis identified three distinct androgen metabolomes: a cluster with mainly gonadal-derived androgen excess (GAE; testosterone, dihydrotestosterone; 21.5% of women), a cluster with mainly adrenal-derived androgen excess (AAE; 11-oxygenated androgens; 21.7%), and a cluster with comparably mild androgen excess (MAE; 56.8%). Age and BMI were similar between groups. Compared to GAE and MAE, the AAE cluster had the highest rates of hirsutism (76.4% vs 67.6% vs 59.9%) and female pattern hair loss (32.1% vs 14.3% vs 21.7%). The AAE cluster had significantly increased insulin resistance as indicated by higher fasting insulin,120min insulin and HOMA-IR, and lower ISI than GAE and MAE clusters (all p&amp;lt;0.01). The AAE cluster had a 2-3fold higher prevalence of impaired glucose tolerance and newly diagnosed type 2 diabetes. Conclusion: Unsupervised cluster analysis revealed three distinct androgen excess subtypes in PCOS. Women within the adrenal androgen excess cluster had a significantly higher prevalence of insulin resistance, impaired glucose tolerance and type 2 diabetes. These results implicate 11-oxygenated androgens as drivers of metabolic risk in PCOS and provide proof-of-principle for an androgen-based stratification tool to guide preventative and therapeutic strategies in PCOS. Presentation Date: Saturday, June 17, 2023

GWOKM: A novel hybrid optimization algorithm for geochemical anomaly detection based on Grey wolf optimizer and K-means clustering

Identifying the geochemical signatures of valuable mineral deposits using regional geochemical data from stream sediments is a challenging task due to the intricate characteristics of geological formations. Our team is currently investigating the potential of unsupervised clustering analysis (CA) and hybridization with the grey wolf optimizer (GWO) in developing multi-element geochemical models using stream sediment data. To cluster the geochemical data and uncover any unusual patterns, we opted to use the K-means (KM) algorithm due to its straightforward implementation, fast computation speed, and capacity to handle the large datasets. Despite its benefits, the KM method also has notable limitations, such as the random selection of cluster centroids. This can result in higher systematic uncertainty in unsupervised geochemical modeling and longer computation times. To mitigate this concern, we have introduced a new hybrid approach, grey wolf optimizer with K-means so-called the GWOKM algorithm to enhance the delineation of multi-elemental patterns in stream sediment geochemical data. This method incorporates the grey wolf optimization algorithm with KM to optimize the identification of both anomalies and backgrounds using factor analysis and sample catchment basin modeling techniques. This approach was utilized to detect anomalous multi-elemental geochemical patterns indicative of porphyry and skarn copper deposits in the Baft area, Kerman belt, Iran. Upon comparison of the geochemical models derived from KM and GWOKM clustering methods, the latter outperformed the former, as evidenced by its higher prediction rate. The outcomes affirm the efficacy of unsupervised KM clustering analysis (CA) as a means of breaking down geochemical anomaly-background populations. Moreover, the integration of clustering methods with optimization algorithms has proven to be successful for enhancing the credibility of mineralized areas, which could be advantageous in future exploration phases. Based on the results, the GWOKM approach generates more reliable and efficient geochemical anomaly targets.

Unsupervised Ship Detection in SAR Imagery Based on Energy Density-Induced Clustering

Article Unsupervised Ship Detection in SAR Imagery Based on Energy Density-Induced Clustering Zifeng Yuan 1, Yu Li 1,*, Yu Liu 1, Jiale Liang 1, and Yuanzhi Zhang 2,3 1 Faculty of Information Technology, Beijing University of Technology, Beijing 100124, China 2 School of Astronomy and Space Science, University of Chinese Academy of Sciences, Beijing 100049, China 3 Key Laboratory of Lunar and Deep Space Exploration, National Astronomical Observatories, Chinese Academy of Sciences, Beijing 100101, China * Correspondence: yuli@bjut.edu.cn Received: 6 March 2023 Accepted: 24 April 2023 Published: 26 September 2023 Abstract: Intelligent recognition of maritime ship targets from synthetic aperture radar (SAR) imagery is a hot research issue. However, interferences such as the strong sea clutter, sidelobe, small ship size and weak backscattered signal continually affect the detection results. To address this problem, a novel unsupervised machine learning-based ship detection algorithm, named energy density-induced clustering (EDIC), is proposed in this paper. It is discovered that the singular values between ship targets and interference signals are significantly different in a local region because of their various concentration degrees of signal energy intensity. Accordingly, in this study, two novel energy density features are proposed based on the singular value decomposition in order to effectively highlight the ship targets and suppress the interference. The proposed novel energy density features have the advantage of clearly distinguishing ship targets from sea surfaces regardless of the effects of interferences. To test the performance of the proposed features, unsupervised K-means clustering is conducted for obtaining ship detection results. Compared with the classical and state-of-the-art SAR ship detectors, the proposed EDIC method generally yields the best performance in almost all tested sea sample areas with different kinds of interferences, in terms of both detection accuracy and processing efficiency. The proposed energy density-based feature extraction method also has great potential for supervised classification using neural networks, random forests, etc.

OS03.6.A UNSUPERVISED CLUSTERING OF MORPHOLOGY PATTERNS ON WHOLE SLIDE IMAGES GUIDE PROGNOSTIC STRATIFICATION OF GLIOBLASTOMA PATIENTS

Abstract BACKGROUND Glioblastoma is the most common malignant adult brain tumor with a poor prognosis and heterogeneous morphology. Stratifying glioblastoma patients according to overall survival (OS) from H&amp;E-stained histopathology whole slide images (WSI) using advanced computational methods is a challenging task with direct clinical implications. We hypothesize that quantifying morphology patterns present in WSI can yield biomarkers of prognostic relevance contributing to optimizing clinical decision-making. MATERIAL AND METHODS This work is based on 188 glioblastoma (IDH-wt, grade 4) cases identified in the TCGA-GBM and TCGA-LGG, based on the 2021 WHO classification criteria. One Diagnostic WSI from each patient at 10X magnification was selected (ntraining/nhold-out-test =152 /36) and labeled as short (&amp;lt;9 months, n=94) or long (&amp;gt;13 months, n=94) survivors. The WSI is split into non-overlapping 256x256 patches, and extensive patch-level WSI curation is conducted to discard artifactual content, i.e., glass reflections, pen markings, black lines on the slide, and tissue tearing. Each patch is described by a feature vector of 512 dimensions using a pre-trained VGG16. Principal component analysis reduced the dimensionality of these vectors to 236 such that 85% variance is retained. Unsupervised k-means clustering revealed distinct groups of morphology patterns, where the number of clusters is automatically determined based on the rand index and silhouette coefficient. The proportions of these patterns describe the tumor’s spatial heterogeneity and are used to distinguish between short and long survivors using a random forest classifier. RESULTS We identified seven clusters of distinct morphology patterns, including 3 categories of tumor cellularity (low/intermediate/high), necrosis/macrophages, and other prognostically-relevant characteristics. The patient-level short vs long survivor prediction is driven by the proportion of these patterns present in WSI achieving classification accuracy, sensitivity and specificity of 83.83%, 94.44%, and 72.72% respectively on the hold-out test set. The high sensitivity signifies the algorithm’s ability to accurately predict the short survivors which is important in treatment planning and patient management. CONCLUSION The interpretability analysis of various morphologic patterns in distinct histologic sub-regions of glioblastoma may uncover correlations with short and long survivors, providing a deeper understanding of the complex relationship between histologic features and clinical outcomes and shedding light on the heterogeneity of glioblastoma. This association could offer additional prognostic information to clinical neuropathologists during microscopic assessment, leading to more refined prognostication and potentially impacting clinical decision-making for improved patient outcomes.

Application of Unsupervised/Supervised Machine Learning to Harmful Gas Detection/Classification Based on Mutually-Interacting Semiconductor Metal Oxide Gas Sensors

Chemical and/or physical sensors are employed at domestic and industrial sites with the aim to perform a variety of functions, such as prediction, forecasting, prognostics, remaining useful life estimation, anomaly detection, and trend analysis. The sensing information contain the analog signals originating from gas-generated signals, pressure, shear, strain, torsion, temperature, humidity, illumination, electromagnetic radiation, sound, and vibration. Chemical gas sensors can be categorized mainly as semiconductor metal oxide (SMO), electrochemical, or photo-ionization detection (PID) sensors. SMO gas sensors are based on adsorption and desorption of gas molecules onto/from semiconductor metal oxides and however, suffer from the low selectivity to diverse gas species, i.e., the mutually interacting features.Here, we report a high-performance machine learning strategy for gas detection/discrimination against harmful gas combination and adopt the synergistic integration of supervised and unsupervised learning by exploiting a SMO gas sensor array. A 4 SMO sensor array was constructed for detecting carbon monoxide and ethyl alcohol (C2H5OH) mixtures using 15 different gas combinations. Gas detection/discrimination performance was probed in terms of different numbers of gas sensor integrated as an array and machine learning algorithms. Unsupervised K-Means clustering was successfully applied to the rational identification of the similarity features of targeted gases among 4 different groups, which are listed as matrix gas, two single-component gases, and one two-gas by employing only unlabeled voltage-based gas sensing information. Detailed classification was performed through a multitude of supervised algorithms, i.e., 2-layer artificial neural networks (ANNs), 4-layer deep neural networks (DNNs), 1-dimensional convolutional neural networks (1D CNNs), and 2-dimensional convolutional neural networks (2D CNNs). The numerical-based DNNs and image-based CNNs are evaluated to be excellent approaches for gas detection/classification, as corroborated by the highest accuracy and lowest loss parameters.Through the systematic investigation on the influence of the number of sensors on the arrayed gas sensor system, the applicability of machine learning methodology to an arrayed SMO gas sensor system is justified by the four unique features, i.e., i) a data augmentation methodology, ii) machine learning approach of combining K-means clustering and neural networks, and iii) a systematic approach to optimized sensor combinations, potentially recommending the minimized sensor systems based on chemical gas sensors against harmful gas species. Even two SMO sensor combinations are shown to be highly effective in performing gas discrimination against harmful gas species assisted through either numeric-based DNNs or image-based 1D CNNs, overcoming the simple clustering proposed through the unsupervised K-means clustering.

Gene-network analysis predicts clinical response to immunotherapy in patients affected by NSCLC

ObjectivesPredictive biomarkers of response to immune checkpoint inhibitors (ICIs) have been extensively studied in non-small cell lung cancer (NSCLC) with controversial results. Recently, gene-network analysis emerged as a new tool to address tumor biology and behavior, representing a potential tool to evaluate response to therapies. MethodsClinical data and genetic profiles of 644 advanced NSCLCs were retrieved from cBioPortal and the Cancer Genome Atlas (TCGA); 243 ICI-treated NSCLCs were used to identify an immunotherapy response signatures via mutated gene network analysis and K-means unsupervised clustering. Signatures predictive values were tested in an external dataset of 242 cases and assessed versus a control group of 159 NSCLCs treated with standard chemotherapy. ResultsAt least two mutations in the coding sequence of genes belonging to the chromatin remodelling pathway (A signature), and/or at least two mutations of genes involved in cell-to-cell signalling pathways (B signature), showed positive prediction in ICI-treated advanced NSCLC. Signatures performed best when combined for patients undergoing first-line immunotherapy, and for those receiving combined ICIs. ConclusionsAlterations in genes related to chromatin remodelling complexes and cell-to-cell crosstalk may force dysfunctional immune evasion, explaining susceptibility to immunotherapy. Therefore, exploring mutated gene networks could be valuable for determining essential biological interactions, contributing to treatment personalization.

Delineating the heterogeneity of embryo preimplantation development using automated and accurate morphokinetic annotation

PurposeOur objective was to design an automated deep learning model that extracts the morphokinetic events of embryos that were recorded by time-lapse incubators. Using automated annotation, we set out to characterize the temporal heterogeneity of preimplantation development across a large number of embryos.MethodsTo perform a retrospective study, we used a dataset of video files of 67,707 embryos from four IVF clinics. A convolutional neural network (CNN) model was trained to assess the developmental states that appear in single frames from 20,253 manually-annotated embryos. Probability-weighted superposition of multiple predicted states was permitted, thus accounting for visual uncertainties. Superimposed embryo states were collapsed onto discrete series of morphokinetic events via monotonic regression of whole-embryo profiles. Unsupervised K-means clustering was applied to define subpopulations of embryos of distinctive morphokinetic profiles.ResultsWe perform automated assessment of single-frame embryo states with 97% accuracy and demonstrate whole-embryo morphokinetic annotation with R-square 0.994. High quality embryos that had been valid candidates for transfer were clustered into nine subpopulations, as characterized by distinctive developmental dynamics. Retrospective comparative analysis of transfer versus implantation rates reveals differences between embryo clusters as marked by poor synchronization of the third mitotic cell-cleavage cycle.ConclusionsBy demonstrating fully automated, accurate, and standardized morphokinetic annotation of time-lapse embryo recordings from IVF clinics, we provide practical means to overcome current limitations that hinder the implementation of morphokinetic decision-support tools within clinical IVF settings due to inter-observer and intra-observer manual annotation variations and workload constrains. Furthermore, our work provides a platform to address embryo heterogeneity using dimensionality-reduced morphokinetic descriptions of preimplantation development.

The thrombus proteome in stroke reveals a key role of the innate immune system and new insights associated with its etiology, severity, and prognosis

BackgroundNowadays little is known about the molecular profile of the occluding thrombus of patients with ischemic stroke. ObjectivesTo analyze the proteomic profile of thrombi in patients who experienced an ischemic stroke in order to gain insights into disease pathogenesis. MethodsThrombi from an exploratory cohort of patients who experienced a stroke were obtained by thrombectomy and analyzed by sequential window acquisition of all theoretical spectra-mass spectrometry. Unsupervised k-means clustering analysis was performed to stratify patients who experienced a stroke. The proteomic profile was associated with both the neurological function (National Institute of Health Stroke Scale [NIHSS]) and the cerebral involvement (Alberta Stroke Program Early CT Score [ASPECTS]) prior to thrombectomy and the clinical status of patients at 3 months using the modified Rankin Scale. In an independent cohort of 210 patients who experienced a stroke, the potential role of neutrophils in stroke severity was interrogated. ResultsProteomic analysis identified 580 proteins in thrombi, which were stratified into 4 groups: hemostasis, proteasome and neurological diseases, structural proteins, and innate immune system and neutrophils. The thrombus proteome identified 3 clusters of patients with distinctive severity, prognosis, and etiology of the stroke. A protein signature clearly distinguished atherothrombotic and cardioembolic strokes. Several proteins were significantly correlated with the severity of the stroke (NIHSS and ASPECTS). Functional proteomic analysis highlighted the prominent role of neutrophils in stroke severity. This was in line with the association of neutrophil activation markers and count with NIHSS, ASPECTS, and the modified Rankin Scale score 90 days after the event. ConclusionThe use of sequential window acquisition of all theoretical spectra-mass spectrometry in thrombi from patients who experienced an ischemic stroke has provided new insights into pathways and players involved in its etiology, severity, and prognosis. The prominent role of the innate immune system identified might pave the way for the development of new biomarkers and therapeutic approaches in this disease.