Browsing by Subject "Hierarchical clustering"

Now showing 1 - 4 of 4

Open Access
Anomaly detection with sparse unmixing and gaussian mixture modeling of hyperspectral images
(2015-07) Erdinç, Acar
One of the main applications of hyperspectral image analysis is anomaly detection where the problem of interest is the detection of small rare objects that stand out from their surroundings. A common approach to anomaly detection is to rst model the background scene and then to use a detector that quanti es the di erence of a particular pixel from this background. However, identifying the dominant background components and modeling them is a challenging task. We propose an anomaly detection framework that uses Gaussian mixture models for characterizing the scene background in hyperspectral images. First, the full spectrum is divided into several contiguous band groups for dimensionality reduction as well as for exploiting the peculiarities of di erent parts of the spectrum. Then, sparse spectral unmixing is performed for each band group for identifying signi cant endmembers in the scene. Three methods for identifying the dominant background groups such as thresholding, hierarchical clustering and biclustering are used in the endmember abundance space to retrieve the sets of pixel groups that represent dominant background components. Next, these pixel groups are used for initializing individual Gaussian mixture models that are estimated separately for each spectral band group. The proposed method enables automatic identi cation of the number of mixture components and e ective initialization of the estimation procedure for the mixture model. Finally, the Gaussian mixture models for all groups are statistically fused for obtaining the nal anomaly map for the scene. Comparative experiments showed that the proposed methods performed better than two other density-based anomaly detectors, especially for small false positive rates, on an airborne hyperspectral data set.
Open Access
High-frequency return and volatility spillovers among cryptocurrencies
(Routledge, 2021-03-22) Şensoy, Ahmet; Silva, T. C.; Corbet, S.; Tabak, B. M.
We examine the high-frequency return and volatility of major cryptocurrencies and reveal that spillovers among them exist. Our analysis shows that return and volatility clustering structures are distinct among different cryptocurrencies, suggesting that return and volatility might have different spillover patterns. Further investigation via minimal spanning trees points out that BTC, LTC and ETH are the most relevant cryptocurrencies in general, serving as connection hubs for linking many other cryptocurrencies. However, their role is challenged lately, potentially due to the increased usage of other cryptocurrencies in time.
Open Access
Survival analysis and its applications in identifying genes, signatures, and pathways in human cancers
(2021-09) Özhan, Ayşe
Cancer literature makes use of survival analyses focused on gene expression based on univariable or multivariable regression. However, there is still a need to understand whether a) incorporating exon or isoform information on expression would improve estimation of survival in cancer patients; and b) applying multivariable regression to gene sets would allow to obtain cancer-specific independent gene signatures in cancer. Differential usage of individual exons, as well as transcripts, are phenomena common to cancerous tissue when compared to normal tissue. The glioblastoma, GBM; liver cancer LIHC; stomach adenocarcinoma, STAD; and breast carcinoma, BRCA datasets from The Cancer Genome Atlas (TCGA) were investigated to identify individual exons and transcripts with transcriptome-wide impact and significance on survival. Aggregation analyses of exons revealed the important genes for survival in each dataset, including GNA12 in STAD, AKAP13 in LIHC and RBMXL1 and CARS1 in BRCA. GSEA was applied on gene sets formed from the exon-based analysis, revealing distinct enrichment profiles for each dataset as well as overlaps for certain GO terms and KEGG pathways. In the second focus of this thesis, multivariable analyses on gene sets whose expressions were obtained from UCSC Xena were used to create two Shiny applications: one for dataset-specific analyses and one for analyses across TCGA-PANCAN. The dataset specific SmulTCan application incorporates Cox regression analyses with expressions of input genes of the user’s choice. The SmulTCan application contains additional model validation, best subset selection and prognostic analyses. The ClusterHR application performs clustering analyses with Cox regression results, while it can also be used for bicluster identification and comparison. The axon-guidance ligand-receptor gene sets Slit-Robo, netrins-receptors and Semas-receptors were used for demonstrating the apps. Several hazard ratio signatures and best subsets that can differentiate between prognostic outcomes have been identified from the input gene sets, as well as ligand-receptor pairs with prognostic significance.
Open Access
Unsupervised segmentation and classification of cervical cell images
(Elsevier BV, 2012-12) Gençtav, A.; Aksoy, S.; Önder, S.
The Pap smear test is a manual screening procedure that is used to detect precancerous changes in cervical cells based on color and shape properties of their nuclei and cytoplasms. Automating this procedure is still an open problem due to the complexities of cell structures. In this paper, we propose an unsupervised approach for the segmentation and classification of cervical cells. The segmentation process involves automatic thresholding to separate the cell regions from the background, a multi-scale hierarchical segmentation algorithm to partition these regions based on homogeneity and circularity, a binary classifier to finalize the separation of nuclei from cytoplasm within the cell regions. Classification is posed as a grouping problem by ranking the cells based on their feature characteristics modeling abnormality degrees. The proposed procedure constructs a tree using hierarchical clustering, then arranges the cells in a linear order by using an optimal leaf ordering algorithm that maximizes the similarity of adjacent leaves without any requirement for training examples or parameter adjustment. Performance evaluation using two data sets show the effectiveness of the proposed approach in images having inconsistent staining, poor contrast, overlapping cells. © 2012 Elsevier Ltd.