**6. Conclusion**

18 Will-be-set-by-IN-TECH

Fig. 8. Block execution time of each node compared to the same block executed in sequential

the main input into smaller chunks and ran the sequential execution of individual blocks: the total processing time for 67 blocks on wn2 was equal to 6h 56min 51s. In the second test the splitting was delegated to the Master Node, which has also been responsible for the distribution of the single blocks to different nodes. The execution time of the build process of the blocks and the file transfer is 2min 55s, and therefore it is negligible if compared to the total execution time. The execution times of different nodes are very similar, only the virtual machine wn5 has slightly better performance, probably due to the higher performance processor. The total execution time of the grid is equal to the maximum execution time of individual nodes, i.e., 1h 42min 19s of wn1. Table 2 summarizes the execution times. Figure 7 depicts total execution time comparison between grid environment and sequential execution on different nodes, on the grid and on sequential execution. The total time reduction is of 75%. Figure 7 shows the comparison between the processing times of individual nodes on the grid and the sequential execution: the colors of the column wn2-seq, correspond to the processing time of each node, through this comparison it can be appreciate the performance gain of the grid. In Figure 8 the comparison between the execution of individual blocks on the grid and in sequential mode is represented. The columns are divided into small parts that correspond to different executed blocks, each block is identified by a different color. The graph is useful for reasoning on the overhead introduced both by the grid and by the virtual machines. From the comparison between the execution of a group of blocks on the grid on a given node and sequentially on the same machine (i.e., the second pair of columns) it can be deduced that the grid introduces an overhead (equal to 2min 38s in this case) due to the transfer of files output,

More and more applications require high performance computing, flexibility and reduced processing time. The architecture explained before, can be useful in many fields i.e., in

mode (wn2-seq)

but negligible compared to the total execution time.

**5.1 Application domains overview**

The solution for CEM problems requires infrastructures based on high performance computing in order to reduce the execution time. However, it is not always possible to use supercomputers or parallelize algorithms code used for the calculation, a good solution may be represented by the Domain Decomposition technique, which allows the subdivision of the original complex problem into a set of smaller subproblems in a grid environment. In fact for this project, a grid infrastructure was carried out, that consists of 6 nodes (both physical and virtual). The results showed that the adoption of this infrastructure has reduced the execution time of 75% with respect to the sequential execution on a single machine. It was also noted that the overhead introduced by the grid is negligible when compared to the total execution time. The Virtualization has allowed optimizing the hardware resources of the machines, letting to run multiple blocks in parallel on the same physical machine, not introducing a significant overhead compared to the overall system performance. Studies are underway to improve the implementation of the scheduler, ensuring that blocks will be distributed to nodes most suitable for their execution. In particular, it will be developed a system that takes into account the weights of the jobs and the weights of the nodes available, assigned on the basis of predetermined criteria (e.g., file size, hardware resources, nodes availability, the average time of execution). The Scheduler will also be able to turn on and off virtual nodes according to the needs of the system, considering the load level of the grid. Furthermore the virtual nodes will be configured dynamically as needed to perform specific jobs (RAM, CPU and storage). A further improvement will be the integration of the system with Public Cloud Platforms (e.g., Amazon EC2) in order to increase the system scalability, asking for virtual nodes usage only when necessary.

#### **7. References**

Andriulli, F., Vipiana, F. & Vecchi, G. (2008). *Hierarchical bases for non-hierarchic 3D triangular meshes*, IEEE Trans. on Antennas and Propagation.

Asadzadeh, P., Buyya, R., Kei, C. L., Nayar, D. & Venugopal, S. (2005). *Global Grids and Software Toolkits: A Study of Four Grid Middleware Technologies*, available: http://www.buyya.com/papers/gmchapter.pdf.

**12** 

*Malaysia* 

**Characterization of Hepatic Lesions** 

*School of Science and Technology, Wawasan Open University, Penang,* 

Magnetic Resonance Imaging (MRI) images have been widely used for liver disease diagnosis. Designing and developing computer-assisted image processing techniques to help doctors improve their diagnosis has received considerable interest over the past years. In this paper, a computer-aided diagnostic (CAD) system for the characterization of hepatic lesions, specifically cyst and tumor as well as healthy liver, from MRI images using texture features and implementation of grid computing (Globus approach) and neural networks (NN) is presented. Texture analysis is used to determine the changes in functional characteristics of organs at the onset of a liver disease, Region of interest (ROI) extracted from MRI images are used as the input to characterize different tissue, namely liver cyst and healthy liver using first-order statistics. The results for first-order statistics are given and their potential applicability in grid computing is discussed. The measurements extracted from First-order statistic include entropy and correlation achieved obvious classification

In this chapter, texture analysis of liver MRI images based on the Spatial Grey Level Cooccurrence Matrix (SGLCM) [3] is proposed to discriminate normal, malignant hepatic tissue (i.e. liver tumor) and cysts in MRI images of the abdomen. SGLCM, also known as Grey Tone Spatial Dependency Matrix [3], is a tabulation of how often different combinations of pixel brightness values (i.e. grey-level) occur in an image. Regions of interest (ROI) from cysts, tumor and healthy liver were used as input for the SGLCM calculation. Second order statistical texture features estimated from the SGLCM are then applied to a Feed-forward Neural Network (FNN) and Globus toolkit for the characterization of suspected liver tissue from MRI images for hepatic lesions classification. This project proposed an automated distributed processing framework for high-throughput, large-scale applications targeted for characterization of liver texture statistical measurements mainly healthy liver, fatty liver, liver

Table 1 lists eight second-order statistical calculations based on SGLCM, namely, contrast, entropy, correlation, homogeneity, cluster tendency, inverse difference moment, energy, and angular second moment, which have shown useful results in hepatic lesions classification for liver tumor using Computed Tomography (CT), Ultrasonography (US) and

**1. Introduction** 

range in detecting different tissues in this work.

cyst for MRI (Magnetic Resonance Imaging) images.

**Using Grid Computing (Globus)** 

Sheng Hung Chung1 and Ean Teng Khor2

**and Neural Networks** 

	- http://globus.org/toolkit/ docs/4.0/security/cas.
	- http://www.gridforum.org/About/abt\_introduction.php.
	- http://searchsoa.techtarget.com/definition/Open-Grid-Services-Architecture.
	- http://openmp.org/wp/.
