Original scientific paper

EXTREME LEARNING MACHINE FOR CLASSIFICATION OF BRAIN TUMOR IN 3D MR IMAGES

S.N. Deepa ; Anna University, Coimbatore, India
B. Arunadevi ; Anna University, Coimbatore, India

Abstract

Extreme Learning machine (ELM) a widely adopted algorithm in machine learning field is proposed for the use of pattern classification model using 3D MRI images for identifying tissue abnormalities in brain histology. The four class classification includes gray matter, white matter, cerebrospinal-fluid and tumor. The 3D MRI assessed by a pathologist indicates the ROI and the images are normalized. Texture features for each of the sub-regions is based on the Run-length Matrix, Co-occurence Matrix, Intensity, Euclidean distance, Gradient vector and neighbourhood statistics. Genetic Algorithm is custom designed to extract and sub-select a decisive optimal bank of features which are then used to model the ELM classifier and best selection of ELM algorithm parameters to handle sparse image data. The algorithm is explored using different activation function and the effect of number of neurons in the hidden layer by using different ratios of the number of features in the training and test data. The ELM classification outperformed in terms of accuracy, sensitivity and specificity as 93.20 %, 91.6 %, and 97.98% for discrimination of brain and pathological tumor tissue classification against state-of-the-art feature extraction methods and classifiers in the literature for publicly available SPL dataset.

Keywords

Brain Tumor; Genetic Algorithm; Extreme Learning Machine

Hrčak ID:

106430

URI

https://hrcak.srce.hr/106430

Publication date:

30.6.2013.

Article data in other languages: croatian

Visits: 3.993 *