Optimizing 3D Brain Tumor Detection with Hybrid Mean Clustering and Ensemble Classifiers

Abstract

Magnetic Resonance Imaging (MRI) scans are extensively used in the medical field for identifying brain cancers. However, a major drawback is that accurate diagnosis of brain tumours can be difficult. This happens due to several reasons such as noise oversegmented regions with a big feature space and high false-positive error rate. This paper presents a novel hybrid algorithm for brain tumour detection. The method exploits state of the art data-preprocessing, feature selection, clustering and classification techniques. The proposed method uses ensemble feature selection to select the most relevant features and ultimately increase the predictive accuracy of the detection system. A hybrid means clustering approach with modulus function clusters the similar data points, filter out noise and identify patterns and anomalies related to brain tumours. The ensemble classifier further explores the accuracy of the proposed scheme and helps in improving the predictive performance of brad tumours. Use of feature selection can highly boost the detection and finally achieve accurate diagnosis and prediction. Eventually, brad tumors can be treated at an earlier stage without being fully developed. The experimental outcomes are promising to the extent that the proposed algorithm can be further incorporated into the treatment of brad tumor. The parameters can be tuned for better performance of the algorithm.