|
doi:10.3850/978-981-08-7618-0_0328
Feature Extraction and Selection for Medical Images- A Hybrid Approach
M. Vasantha1 and V. Subbiah Bharathy2
1Research Scholar, Mother Teresa University, Kodaikanal, India.
2DMI College of Engineering, Chennai, India.
ABSTRACT
Digital mammography plays an important role in Computer Aided Detection of Breast cancer. A common approach to improve medical image classification is to extract and select the relevant features for the classification. Considering all extracted features for categorization is not always beneficial. The number of commonly used features in the literature for training of image classification is many. Existing algorithms for selecting a subset of available features for image analysis fail to adequately eliminate redundant features. This paper presents a new selection algorithm based on a hybrid approach. It uses Sequential Search Forward (SFS) and Genetic Algorithm to select relevant feature for classifier decision making. A database of 113 mammograms from the Mammographic Image Analysis Society (MIAS) was used for our experiments. For classification of samples, we have employed the freely available Machine Learning package, WEKA to train our data set using J48 decision tree method. Out of 113 images in the dataset, 80 were used for training and the remaining for testing purposes. Cancer detection accuracy using a feature set selected by our algorithm produce substantially similar accuracy as using a
26- feature set selected by SFS alone. We also observed that this method greatly reduce the amount of computation needed for classification.
Keywords: Digital mammography, Genetic algorithm, Mammographic image analysis society.
Back to TOC
FULL TEXT(PDF)
|