K. Thangavel

ELM-Based Ensemble Classifier for Gas Sensor Array Drift Dataset

Much work has been done on classification for the past fifteen years to develop adapted techniques and robust algorithms. The problem of data correction in the presence of simultaneous sources of drift, other than sensor drift, should also be investigated, since it is often the case in practical situations. ELM is a competitive machine learning technique, which has been applied in different domains for classification. In this paper, ELM with different activation functions has been implemented for gas sensor array drift dataset. The experimental results show that the ELM with bipolar function classifies the drift dataset with an average accuracy of 96 % than the other function. The proposed method is compared with SVM.

Mammogram image segmentation using fuzzy clustering

This paper proposes mammogram image segmentation using Fuzzy C-Means (FCM) clustering algorithm. The median filter is used for pre-processing of image. It is normally used to reduce noise in an image. The 14 Haralick features are extracted from mammogram image using Gray Level Co-occurrence Matrix (GLCM) for different angles. The features are clustered by K-Means and FCM algorithms inorder to segment the region of interests for further classification. The performance of segmentation result of the proposed algorithm is measured according to the error values such as Mean Square Error (MSE) and Root Means Square Error (RMSE). The Mammogram images used in our experiment are obtained from MIAS database.

A novel approach to select significant genes of leukemia cancer data using K-Means clustering

DNA microarray technologies are leading to an explosion in available gene expression data which simultaneously monitor the expression pattern of thousands of genes. All the genes may not be biologically significant in diagnosing the disease. In this paper, a novel approach has been proposed to select significant genes of leukemia cancer using K-Means clustering algorithm. It is an unsupervised machine learning approach, which is being used to identify the unknown patterns from the huge amount of data. The proposed K-Means algorithm has been experimented to cluster the genes for K=5,10 and 15. The significant genes have been identified through the best accuracy obtained from the clusters generated. The accuracy of the clusters are determined again by using K-Means algorithm compared with ground truth values.

Mammogram image segmentation using granular computing based on rough entropy

The mammography is the most effective procedure for to diagnosis the breast cancer at an early stage. A granule is a mass of objects, in the universe of discourse, put together by indistinguishability, similarity, proximity, or functionality. In mammograms, it is quite difficult to identify the suspicious region which is a mass of calcification on the breast tissue. This paper proposes rough entropy based granular computing to segment mammogram images. The proposed method is evaluated by classification algorithms which are available in WEKA.

Computed radiography skull image enhancement using Wiener filter

Medical imaging devices are used to scan different organs of human being and used in different stages of analysis. Magnetic Resonance Image (MRI), Computer Tomography (CT), Ultrasound and X-Ray are some of the imaging techniques adopted for acquiring images to diagnose most of the diseases. The main aim of this study is to improve the quality of Computed Radiography (CR) medical images. Denoising with edge preservation is very important in CR X-Ray imaging. Noise reduction should be a great concern in order not to lose detailed spatial information for perfect and optimal diagnosis of diseases. Computing techniques also need to be taken care of since the digital format of the medical images is comprised with large sized matrices. In this study, firstly, we compared a series of filtering techniques using Wiener filtering method to remove the Poisson noise from CR X-Ray human Skull images. Secondly, Contrast Enhancement was performed by using Histogram Equalization and intensity value adjustment with limits points. The main aim of this work is to improve the visual quality of CR X-Ray human skull images and enhance the subtle details such as edges and nodules, which are with low contrast white circular objects. The performance of the proposed method is analyzed using Means Square Error (MSE) and Peak Signal Noise Ratio (PSNR) measures. Experimental results show that Wiener Filtering method effectively reduce the Poisson noise from CR X-Ray of a human Skull image. Finally the study is concluded with future implications for research areas.

A review of early detection of cancers using breath analysis

Authentic and accurate information is basic to any disease control initiative. More than 70% of diseases are related to life-style factors such as food and beverage practices, personal habits, infections, tobacco consumption and social customs. In addition, urbanization, industrialization and increasing life-span are also known to influence the cancer pattern globally. This necessitates proper appreciation of risk factors and other causes of cancer by the people. Various modalities for early detection through screening are being investigated. Majority of the patients have locally advanced or disseminated disease at presentation and are not candidates for surgery. Chemotherapy applied as an adjunct with radiation improves survival and the quality of life. New anticancer drugs, which have emerged during the last decade, have shown an improved efficacy toxicity ratio. This review is more about the diagnosing cancer at an early stage using invasive electronic sensors and intelligent computing methods by capturing only the breath of the human being. Strengthening the methods for early diagnosis of cancers and improved treatments will have a significant impact on cutting death rates.

Protein sequence motif patterns using adaptive Fuzzy C-Means granular computing model

Data Mining is the process to extract hidden predictive information from large databases. In Bioinformatics, data mining enables researchers to meet the challenge of mining large amount of biomolecular data to discover real knowledge. Major research efforts done in the area of bioinformatics involves sequence analysis, protein structure prediction and gene finding. Proteins are said to be prominent molecules in our cells. They involve virtually in all cell functions. The activities and functions of proteins can be determined by protein sequence motifs. These protein motifs are identified from the segments of protein sequences. All segments may not be important to produce good motif patterns. The generated sequence segments do not have classes or labels. Hence, unsupervised segment selection technique is adopted to select significant segments. Therefore Singular Value Decomposition (SVD) entropy method is adopted to select significant sequence segments. In this proposed work, weighted K-Means and Adaptive Fuzzy C-Means have been applied to the selected segments to generate granules, since large amount of segments cannot be grouped or clustered as such. Each granules generated by weighted K-Means algorithm are further clustered by using the K-Means algorithm and granules generated by Adaptive Fuzzy C-Means algorithm are clustered by using Weighted K-Means. The two proposed models are compared with K-Means granular computing model. The experimental results show that Adaptive Fuzzy C-Means with Weighted K-Means technique produces better results than K-Means and weighted K-Means granular computing methods.

Outliers detection on protein localization sites by partitional clustering methods

A large molecule composed of one or more chains of amino acids in a specific order, the order is determined by the base sequence of nucleotides in the gene that codes for the protein. Proteins are required for the structure, function, and regulation of the bodys cells, tissues, and organs and each protein has unique functions. Localization sites of proteins are identified by the mechanism and moved to its corresponding organelles. In this paper, we introduce the method clustering and its types K-Means and K-Medoids. The clustering algorithms are improved by implementing the two initial centroid selection methods instead of selecting centroid randomly. K-Means algorithm can be improved by implementing the initial cluster centroids are selected by the two proposed algorithms instead of selecting centroids randomly, which is compared by using Davie Bouldin index measure, hence the proposed algorithm1 overcomes the drawbacks of selecting initial cluster centers then other methods. In the yeast dataset, the defective proteins (objects) are considered as outliers, which are identified by the clustering methods with ADOC (Average Distance between Object and Centroid) function. The outliers detection method and performance analysis method are studied and compared, the experimental results shows that the K-Medoids method performs well when compare with the K-Means clustering.

Empirical study on early detection of lung cancer using breath analysis

A variety of modalities for early detection of cancer through screening are being investigated. The main inspiration from electronic noses is the development of qualitative, low-cost, real-time, and portable methods to perform reliable, objective, and reproducible measures of volatile compounds and odors. To defeat the difficulties faced by the existing clinical diagnostic and detection methods, the breathe analysis is proposed. Breathe analysis is done using an electronic nose. Breathe from the humans are sampled and the Volatile Organic Compounds (VOC) present in the sampled breathe are analyzed to identify the existing component pertaining to the particular cancer. In this paper, an empirical study on early detection of cancer is achieved by build the electronic nose. The proposed electronic nose has been implemented by Backpropagation Neural Network (BPN).

Mammogram image feature selection using unsupervised tolerance rough set relative reduct algorithm

Feature Selection (FS) aims to determine a minimal feature subset from a problem domain while retaining a suitably high accuracy in representing the original features. Rough set theory (RST) has been used as such a tool with much success. In the supervised FS methods, various feature subsets are evaluated using an evaluation function or metric to select only those features which are related to the decision classes of the data under consideration. However, for many data mining applications, decision class labels are often unknown or incomplete, thus indicating the significance of unsupervised feature selection. However, in unsupervised learning, decision class labels are not provided. The problem is that not all features are important. Some of the features may be redundant, and others may be irrelevant and noisy. In this paper, a novel unsupervised feature selection in mammogram image, using tolerance rough set based relative reduct is proposed. And also, compared with Tolerance Quick Reduct and PSO - Relative Reduct unsupervised feature selection methods. A typical mammogram image processing system generally consists of mammogram image acquisition, pre-processing of image segmentation, feature extraction, feature selection and classification. The proposed method is used to reduce features from the extracted features and the method is compared with existing unsupervised features selection methods. The proposed method is evaluated through clustering and classification algorithms in K-means and WEKA.