Automated Multi-class Brain Tumor Types Detection by Extracting RICA Based Features and Employing Machine Learning Techniques

Brain tumor is the leading reason of mortality across the globe. It is obvious that the chances of survival can be increased if the tumor is identified and properly classified at an initial stage. Several factors such as type, texture and location help to categorize the brain tumor. In this study, we extracted reconstruction independent component analysis (RICA) base features from brain tumor types such as glioma, meningioma, pituitary and applied robust machine learning algorithms such as linear discriminant analysis (LDA) and support vector machine (SVM) with linear and quadratic kernels. The jackknife 10-fold cross validation was used for training and testing data validation. The SVM with quadratic kernel gives the highest multiclass detection performance. To detect pituitary, the highest detection performance was obtained with sensitivity (93.85%), specificity (100%), PPV (100%), NPV (97.27%), accuracy (98.07%) and AUC (96.92). To detect glioma, the highest detection performance was obtained with accuracy (94.35%), AUC (0.9508). To detect the meningioma, the highest was obtained with accuracy (96.18%), AUC (0.9095). The findings reveal that proposed methodology based on RICA features to detect multiclass brain tumor types will be very useful for treatment modification to achieve better clinical outcomes.