T. S. Subashini

Automated assessment of breast tissue density in digital mammograms

Mammographic density is known to be an important indicator of breast cancer risk. Classification of mammographic density based on statistical features has been investigated previously. However, in those approaches the entire breast including the pectoral muscle has been processed to extract features. In this approach the region of interest is restricted to the breast tissue alone eliminating the artifacts, background and the pectoral muscle. The mammogram images used in this study are from the Mini-MIAS digital database. Here, we describe the development of an automatic breast tissue classification methodology, which can be summarized in a number of distinct steps: (1) preprocessing, (2) feature extraction, and (3) classification. Gray level thresholding and connected component labeling is used to eliminate the artifacts and pectoral muscles from the region of interest. Statistical features are extracted from this region which signify the important texture features of breast tissue. These features are fed to the support vector machine (SVM) classifier to classify it into any of the three classes namely fatty, glandular and dense tissue.The classifier accuracy obtained is 95.44%.

Breast mass classification based on cytological patterns using RBFNN and SVM

Correct diagnosis is one of the major problems in medical field. This includes the limitation of human expertise in diagnosing the disease manually. From the literature it has been found that pattern classification techniques such as support vector machines (SVM) and radial basis function neural network (RBFNN) can help them to improve in this domain. RBFNN and SVM with their remarkable ability to derive meaning from complicated or imprecise data, can be used to extract patterns and detect trends that are too complex to be noticed by either humans or other computer techniques. This paper compares the use of polynomial kernel of SVM and RBFNN in ascertaining the diagnostic accuracy of cytological data obtained from the Wisconsin breast cancer database. The data set includes nine different attributes and two categories of tumors namely benign and malignant. Known sets of cytologically proven tumor data was used to train the models to categorize cancer patients according to their diagnosis. Performance measures such as accuracy, specificity, sensitivity, F-score and other metrics used in medical diagnosis such as Youdens index and discriminant power were evaluated to convey and compare the qualities of the classifiers. This research has demonstrated that RBFNN outperformed the polynomial kernel of SVM for correctly classifying the tumors.

Detection of Heart Muscle Damage from Automated Analysis of Echocardiogram Video

ABSTRACT In this work, an approach for heart muscle damage detection from echocardiography sequences is proposed. To exemplify the approach, a system is presented which involves image denoising and enhancement and segmentation of the left ventricle (LV) for extracting the heart wall boundaries. Using the heart wall boundaries global LV parameters are calculated followed by statistical pattern recognition and classification to identify the heart muscle damage or myocardial ischemia (MI). The performance of this algorithm is assessed in 60 real patient data with both normal and abnormal conditions. The experimental results reveal that the proposed method can be used as an effective tool for detection of heart muscle damage or MI automatically.

Static Hand Gesture Recognition for Sign Language Alphabets using Edge Oriented Histogram and Multi Class SVM

In recent years, enormous research is progressing in the field of Computer Vision and Human Computer Interaction where hand gestures play a vital role. Hand gestures are more powerful means of communication for hearing impaired when they communicate to the normal people everywhere in day to day life. As the normal people find little difficulty in recognizing and interpreting the meaning of sign language expressed by the hearing impaired, it is inevitable to have an interpreter for translation of sign language. To overcome this difficulty, an automatic hand gesture recognition system which translates the sign language into text needs to be developed. In this paper, a static hand gesture recognition system for American Sign Language using Edge Oriented Histogram (EOH) features and multiclass SVM is proposed. The edge histogram count of input sign language alphabets is extracted as the features and applied to a multiclass SVM for classification. The average accuracy of the system is compared with different number of features and the experimental findings demonstrate that the proposed method gives a success rate of 93.75%.

Pectoral Muscle removal and Detection of masses in Digital Mammogram using CCL

A mammogram is a radiograph of the breast tissue. It is an effective non-invasive means of examining the breast, commonly searching for breast cancer. Cancer is not preventable, but early detection leads to a much higher chance of recovery and lowers the mortality rate. Due to the high volume of images to be analyzed by radiologists, and since senior radiologists are rare, the accuracy rate tends to decrease. This is reflected in the high percentage of unnecessary biopsies that are performed and many deaths caused by late detection or poor diagnosis. This paper proposes a computer aided diagnosis system for detecting masses in mammograms using connected component labeling(CCL). This paper also addresses the problem of eliminating and pectoral muscles from the mammogram before the detection process so that further processing is confined to the breast region alone.

Automatic Identification and Elimination of Pectoral Muscle in Digital Mammograms

aided detection/diagnosis aims at assisting radiologist in the analysis of digital mammograms. Digital mammogram has emerged as the most popular screening technique for early detection of breast cancer and other abnormalities in human breast tissue. The pectoral muscle represents a predominant density region in most mammograms and can affect/bias the results of image processing methods. This paper addresses the problem of eliminating the pectoral muscles from the mammogram so that further processing for detection and diagnosis of breast cancer is confined to the breast region alone. The proposed work is done in three steps. In the first step, the mammogram is oriented to the left to minimize computations. In the second step the top left quadrant of the mammogram which contains the pectoral muscle is extracted. Next, the pectoral muscle contour is computed using our proposed algorithm. Totally 120 mammogram images were taken up for the study. A comprehensive comparison with manually-drawn contours by the radiologist reveals the strength of the proposed method and shows that it can be effectively used as a preprocessing step in the design of CAD system for breast cancer.

Transform based approaches for Palmprint Identification

In this paper, identification of a person is carried out using Discrete Cosine Transform (DCT) and Discrete Wavelet Transform (DWT) coefficients as palmprint features. Dimensionality reduction was carried out using Principal Component Analysis (PCA) and the reduced features were classified using Support Vector Machine (SVM). The segmented palmprint images used in this study are from the IITD palmprint database. Here, we describe the palmprint identification methodology, which can be summarized in a number of distinct steps: (1) preprocessing, (2) feature extraction, and (3) classification. Experiments were developed on a database of 100 images from 20 individuals. This study shows that DCT is better for palmprint identification with an accuracy of 99% compared to DWT which gives 96.66% accuracy.

Detection of Cardiac Abnormality from Measures Calculated from Segmented Left Ventricle in Ultrasound Videos

In this paper a novel and robust automatic LV segmentation by measuring the properties of each connected components in the echocardiogram images and a cardiac abnormality detection method based on ejection fraction is proposed. Starting from echocardiogram videos of normal and abnormal hearts, the left ventricle is first segmented using connected component labeling and from the segmented LV region the proposed algorithm is used to calculate the left ventricle diameter. The diameter derived is used to calculate the various LV parameters. In each heart beat or cardiac cycle, the volumetric fraction of blood pumped out of the left ventricle (LV) and the ejection fraction (EF) were calculated based on which the cardiac abnormality is decided. The proposed method gave an accuracy of 93.3% and it can be used as an effective tool to segment left ventricle boundary and for classifying the heart as either normal or abnormal.

An Approach toward the Efficient Indexing and Retrieval on Medical X-Ray Images

content-based image retrieval (CBIR) has become one of the most active areas of research in computer vision. With rapid advances in digital imaging modalities, the use of CBIR to search for the clinically relevant and visually similar medical images is highly felt nowadays. This paper proposes a system for content based image retrieval of X-ray images.The six classes of X-ray images used for this work are from the IRMA ImageCLEFmed 2008 database. Discrete Cosine Transform (DCT) coefficients were used as features and the X-rays were classified using Support Vector Machine (SVM). The classified images along with the features were stored in the database using hierarchical index structure. Euclidean distance is used as the metric for retrieving the top three images from the database relevant to the given query image

Automatic Detection and Inpainting of Text Images

The proposed system detects text using connect component labelling and a set of selection/ rejection criteria which helps to retain the text region alone. The detected text region is then inpainted using fast marching algorithm which uses the pixel information that is present in the non-text region of the image for inpainting the detected text region. This work is done in two steps. The first step detects the text region from the image without the user manually marking it and in the second step the text is de-occluded from the image using the existing fast marching inpainting algorithm.