Find research datasets worth reusing

Figure 9. Performance analysis of FCM-PSO, GPC-PSO and GFCM-PSO-An Optimized Clustering Approach for Automated Detection of White Matter Lesions in MRI Brain Images

Figure 8. Performance analysis of FCM, GPC and GFCM Figure 9.-An Optimized Clustering Approach for Automated Detection of White Matter Lesions in MRI Brain Images

All scans obtained from different image clustering models are manually ranked based on values in table 1. Table 2 represents WML detection rates of optimized images. FCM, GPC and GFCM clustering methods and hybrid optimized methods (FCM-PSO, GPC-PSO and GFCM-PSO) are applied on a dataset of 208 images and ranking is done in terms of under detected, over detected, properly detected as shown in figure 8 and figure 9. The number of images detected properly in GFCM is comparatively high than FCM and GPC. The optimized result of GFMC provides accurate detection of WMLs and it properly detects 195 images.

Figures -using Particle Swarm Optimization (PSO)-An Optimized Clustering Approach for Automated Detection of White Matter Lesions in MRI Brain Images

The performance of WML quantification is evaluated using clustering algorithms. When the image is pre-processed, contrast of the image is enhanced. The resulting enhanced image is clustered using the effective clustering algorithms. Figure 3 represents the input image for WML detection. In order to increase robustness, the noisy medical image is pre-processed. Figure 4 depicts the pre-processed image. Bright contrast stretching, which is one of the image enhancement (pre-processing) techniques is applied. After pre-processing the enhanced image is subjected to clustering. Three clustering models are proposed to provide accurate results. All scans obtained from different image clustering models are manually ranked based on values in table 1. Table 2 represents WML detection rates of optimized images. FCM, GPC and GFCM clustering methods and hybrid optimized methods (FCM-PSO, GPC-PSO and GFCM-PSO) are applied on a dataset of 208 images and ranking is done in terms of under detected, over detected, properly detected as shown in figure 8 and figure 9. The number of images detected properly in GFCM is comparatively high than FCM and GPC.

Figures a, b ,c -8-An Optimized Clustering Approach for Automated Detection of White Matter Lesions in MRI Brain Images

In turn, the optimized results of GFCM provide overall accuracy of 95%. Figures 8(a), 8(b) and 8(c) shows the comparison results of clustering models and optimization technique in terms of Se, Sp and Acc.

opencc-by-4.0Dec 2014View details →

BRAIN Journal-Electrophysiological Neuroimaging using sLORETA Comparing 100 Schizophrenia Patients to 48 Patients with Major Depression -Figure 4. Histogram age distributions of the 100 Schizophrenia patients illustrating clusters of patients at 20, 30, 40, and 50 years old.

The figure below is a histogram distribution of the ages of one hundred Schizophrenia patients in this study. There appears to be a cyclical peak every ten-years cycles at 20, 30, 40, and 50-year-old patients. This may suggest a recent finding that CD8 T cells, which play a pivotal role in mediating long-term immunity to Toxoplasma, are down-regulated in schizophrenia patients (Bhadra et al., 2013). Based on the aforementioned statements, a vaccine for Toxoplasmosis may be difficult to achieve, but the treatment and screening possibilities in patients with psychosis are available today.

Figure2. Generation of negative feedbacks gets tuned once a TCR completes stimulation beyond the threshold l. A TCell generates activation signal to BCell once it gets stimulation of its k-TCRs.-AIDEN: A Density Conscious Artificial Immune System for Automatic Discovery of Arbitrary Shape Clusters in Spatial Patterns

A TCR at position p is stimulated if rp (x) - rn(x) > l. Figure 1 depicts this process. When a T Cell receives stimulations on more than k receptors, it generates activation signal to a B Cell, as represented in Figure2.

Figure1. Static stimulation of a single TCR- The kinetic proofreading by the receptor on input x Є X forwards the receptor position p toward l. The receptor will generate negative feedback if p > β. The receptor will generate success signal when p== l.-AIDEN: A Density Conscious Artificial Immune System for Automatic Discovery of Arbitrary Shape Clusters in Spatial Patterns

FIGURES 2 a-g -AIDEN: A Density Conscious Artificial Immune System for Automatic Discovery of Arbitrary Shape Clusters in Spatial Patterns

The program was implemented in Matlab and tested with several patterns. The first, dataset1 consisted of 2 patterns each comprised of 100 points falling on two concentric circles of radii 10 and 20 respectively. The second, dataset2 consisted of 3 patterns each of 100 points falling on three concentric circles of radii 10, 15, and 20 respectively. The model was further tested for its capability to find clusters in patterns of open spatial form using dataset3 and dataset4 consisting of 200 and 300 points falling on 2 and 3 concentric semi circles respectively. As shown in the Figure2.a and Figure2.b, the algorithm is capable of determining spatial association of a data point with other data points belonging to its appropriate circle only. The results successfully demonstrated the capability of our model to automatically detect clean clusters of arbitrary shapes in the input data represented in closed spatial form. The model was found even capable of determining clusters of open spatial forms also, as shown in Figure2.c and Figure2.e. However, the output of the algorithm was found affected by the values of the algorithm parameters k and a. In the present experiment, k =8 and a =10 was sufficient for performing correct cluster associations. On the other hand, correct clustering for the dataset2, could be obtained with 10NN estimation i.e. k =10, with a.=15. Moreover setting k =15, with a.=15 was required for dataset4, as clustering error was observed with k =10, with a.=15, as in Figure2.d. Figure2.f and Figure2.g show the correct clustering even in presence of combination of open and closed form of input patterns. In each figure, the first sub-plot shows the original data and the second sub-plot shows the clusters identified by our program.

Figure 3. Different domain-Hybridization of Fuzzy Clustering and Hierarchical Method for Link Discovery

In this paper, we propose a new hybrid algorithm, which combines the features of fuzzy algorithm and hierarchical algorithm. Our algorithm decreases the number of comparisons on link discovery. Using hierarchical algorithm in the first level, the data is divided into two groups. In the second level the worst cluster is determined by matrix memberships and then it split. This stage is repeated until the optimal number of clusters is achieved.Creating typed links, between the entities of different datasets is one of the key challenges on web of data .We presented the clustering approach, which decreases the number of comparisons on link discovery. The results of linking the movies in LinkedMDB to corresponding movies in DBpedia and also linking the places in LinkedGeoData to the places of DBpedia show the it reduces the number of comparisons without loss of recall and precision. Hopefully in the future, we will be able to elevate the proposed method recall to 100 % using the membership matrix.

Figure 1. Workflow of approach-Hybridization of Fuzzy Clustering and Hierarchical Method for Link Discovery

In this section, we present our model in more detail. Figure 1 gives an overview of the workflow. Our approach is organized in two phases: first, the division of data in two clusters; thenthe determination of the worst cluster and splitting. The number of clusters is unknown, but our algorithms can find this parameter based on the complexity of cluster structure.

Figure 2. Decrease comparisons.-Hybridization of Fuzzy Clustering and Hierarchical Method for Link Discovery

Figure3. APCs A1-A4 connected within range of cohesion-factor-threshold form members of one ARB-AIDEN: A Density Conscious Artificial Immune System for Automatic Discovery of Arbitrary Shape Clusters in Spatial Patterns

Figure3 depicts this process. The model with the above specification then effectively detects self or non-self pathogens. In terms of its application to the task of clustering, this interpretation means making the affinities high within clusters and low across clusters. A pathogen corresponding to an outlier would not stimulate a TCR sufficiently and may not form part of any ARB.

opencc-by-4.0Nov 2015View details →

Figure 4. After merging, overview is more transparent. Tens of persons were merged together into clusters in order to clarify the visualization. Firms and persons are recognized based on their icons.-Browsing Semantic Data in Slovakia

The usefulness of such visualization has its key points regarding connections. Thanks to SBR browsing module, we were able to get 22 firm records for “Váhostav” query. Between any 2 companies, connections may be (and often are) not bidirectional, so, in order to navigate through connections, we have refined all 22 records. Although, even being filtered, graph is still complex. And it is possible to further navigate and search for outgoing connections, for example firm “MERLIN TRADE, a.s.” on Fig.4 contains item on “Ján Kato”, which is already included in our graph and connected to “VÁHOSTAV&SK&DEVELOPEMENT” on bottom left side and “VÁHOSTAV&SK, a.s.” in the center. Edge coloring and drawing is helpful with overlapped edges. For methods of visualization, including coloring, we refer to studies of H. Omote and K. Sugiyama (2006), and I. Herman, G. Melanon, and M. S. Marshall (2000)  or our study on graph clutter filtering and connectivity distance (Mojzis & Laclavik, 2014).

opencc-by-4.0Jun 2016View details →

BRAIN Journal-Redesigning a Flexible Material Master Data Application with Language Dependency-Figure 7. Dialog structure of the View Cluster

Field dependency has been generated automatically and as a result the following desired structure has been achieved (figure 7). After populating the database tables with data in different languages with the help of the view cluster, the popup has been adapted in order to support the translation of the tabs and areas. Supplementary internal tables have been defined in the function module POPUP_FLEX, in order to copy data from the translation tables. The corresponding SELECT statements have been embedded in TRY-CATCH blocks, in order to prevent short dumps due to faulty selection processes.

opencc-by-4.0Jun 2016View details →

BRAIN Journal-Redesigning a Flexible Material Master Data Application with Language Dependency-Figure 6. View cluster TAFC

For all these tables and views, table maintenance generators have been created and activated, in order to have the possibility to manage individual datasets in every table and view. The corresponding names of those function groups for the table maintenance generators are the same names as those for the views. The purpose of these maintenance views is only to take care of the input data more efficiently. These views will be used later in the view cluster, which ensures a hierarchical order of the data. Therefore, the maintenance views will also include the predecessor, in order to facilitate linking in the field dependency tab of the view cluster (Swapna, 2007). These three maintenance views are embedded in the following view cluster (figure 6)

BRAIN Journal-A New Challenge for Information Mining-Figure 4. Results of Weka's Clustering

In the Figure 4 we show that, in a particular cluster, attributes are grouped in the "good" attribute of "Expertise with technology” with the "Low" attribute of "Student's performance" together. Thus, we can deduce that the level of Student's Performance is influenced by other factors over "Expertise with technology" of teacher. These factors can be searched inside the cluster, providing useful information to a significant exploration. These aspects are not deducible only by exploration through the portal and, for this reason, the clustering technique allows to user to navigate better during the search.

BRAIN Journal-A New Challenge for Information Mining-Figure 3: Clustered Instances

We tested the algorithm with different values of K, to find the optimal centroids. In general, as you know, there is no method for determining the exact value of K, but an accurate estimate can be obtained, for example, monitoring the value of the sum of squared error (SSE) for some values of k (for example 2, 4, 6, 8, etc.). The SSE is defined as the sum of the squared distance between each member of the cluster and its centroid. Mathematically, we can write (1): ( , ) (1) 1 2     K i c x i i SSE dist x c In our case we estimated in k = 8 the best number of cluster. We obtained the following clustered instances.

BRAIN Journal-Personality Questionnaires as a Basis for Improvement of University Courses in Applied Computer Science and Informatics-Figure 5. Hierarchical clustering by scores across the EPQ–R scales for data about all the participants

The clusters were generated using an implementation of a hierarchical clustering algorithm available in the R environment (R, n.d.). The top three clusters were extracted from a hierarchical cluster tree shown in Figure 5, while the color of data points in the visualization shown in figure 4 was determined based on cluster labels. Hierarchical clusters could be used when investigating which students in the analyzed sample share similar personality traits. This could be especially useful for smaller student groups as the teacher may manually inspect the cluster tree and its leaves, which designate individual students. For instance, there are three students in cluster 3, who are represented within the tree in Figure 5 by identifiers 14, 22, and 24. The students with identifiers 14 and 22 are more closely linked and more similar to each other than to the student with identifier 24.

BRAIN Journal-Personality Questionnaires as a Basis for Improvement of University Courses in Applied Computer Science and Informatics-Figure 4. Radial visualization of scores across the EPQ–R scales for clustered data about all the participants

On the other hand, the division of data points by gender might not be the only useful strategy when visually inspecting the analyzed sample in a coordinate system. Numerous clustering algorithms may be used to determine which data points share similar scores across the EPQ–R scales, i.e., which data points belong to the same cluster of similar entities based on their corresponding EPQ–R scores. A radial visualization in which data points were organized into three clusters is given in Figure 4. Each cluster is marked by a different color: cluster 1 by red, cluster 2 by green, and cluster 3 by blue.