aiMRS: A feature extraction method from MRS signals based on artificial immune algorithms for classification of brain tumours

Abstract

Precise diagnosis of brain tumour by experienced radiologists involves a complex set of processes including magnetic resonance imaging, magnetic resonance spectroscopy (MRS) data and histopathological evaluations. In this study, a new hybrid feature extraction method, called as aiMRS, based on the negative selection algorithm and clonal selection algorithm of artificial immune systems is developed on MRS data for the detection and classification of brain tumours. In the study, differentiation of benign and malignant brain tumours, classification of normal brain tissue and brain tumour, and detection of metastasis and primary brain tumours are performed with high precision using pattern recognition methods based on the proposed aiMRS method. According to the experimental results performed on a large data set created with the MRS data obtained from INTERPRET database, when the proposed feature extraction method applied, classification of normal brain tissue and brain tumours, benign and malignant brain tumours and metastasis and primary brain tumours is achieved with 100, 98.58 and 98.94% accuracy, respectively. These results show that this proposed system can be used as a secondary tool in physicians' decision-making processes for the classification of brain tumours.