Authors : Indah Soesanti, Meidar Hadi, Avizenna and Igi Ardiyanto
Abstract: Magnetic Resonance Imaging (MRI) is a medical technique commonly used by radiologists to visualize organ structures in humans without surgery. Based on histopathological appearance, the World Health Organization (WHO) classifies premier tumors into Low Grade Glioma (LGG) and High Grade Glioma (HGG). The process of selecting a tumor area is usually done manually by a radiologist, the process takes a lot of time and effort. To help provide a second opinion for radiologists in the classification of LGG and HGG brain tumors, a computerized system is needed to process ROI, feature extraction and MRI image classification. This study aims to compare the classification results with the ROI process and without the ROI process. 1000 images in the form of 500 LGG Flair MRI images and 500 MRI images of Flair HGG were processed by determining the ROI of tumor images compared to without the ROI processing being performed. The feature extraction process uses statistical texture histogram equalization method by calculating variance, skewness, kurtosis and GLCM texture using Energy, Contrast, Entropy, Homogeneity, Correlation, SumAverage, Variance, Dissimilarity, Auto Correlation. Finally, the Random Forest model is used to classify LGG and HGG class images and be evaluated by k-fold validation validation with k = 7. The results obtained from the proposed method of accuracy, sensitivity and specificity reached 83.6% accuracy, 80.88% sensitivity and 86.84% specificity. Shows that the method used to classify with ROI results in an increase with an accuracy of 4%, sensitivity increases by 4.46% and a specificity of 3.33%. So that, the results obtained accuracy of 87.6% accuracy, 85.34% sensitivity and 90.17% specificity.
Indah Soesanti, Meidar Hadi, Avizenna and Igi Ardiyanto, 2020. Classification of Brain Tumor MRI Image using Random Forest Algorithm and Multilayers Perceptron. Journal of Engineering and Applied Sciences, 15: 3385-3390.