An Adaboost Support Vector Machine Based Harris Hawks Optimization Algorithm for Intelligent Quotient Estimation from MRI Images

Human intelligence is measured using the Intelligent Quotient (IQ) score which is derived from a range of tests. Due to a lack of a large dataset, IQ was estimated utilizing consistent scanning approaches. Multiple datasets scanned from various locations are integrated using different scanning factors and procedures. As a result, there was a lot of diversity among the datasets. To overcome the aforementioned problems, this work presented a novel technique for IQ assessment based on data from Magnetic Resonance Imaging dataset. The pre-processing step focuses on regions like cerebrospinal fluid, white matter (WM), tissue segmentation, cerebellum removal, and skull stripping. Based on Gray Level Co-occurrence Matrix features, we have extracted energy, entropy, contrast, homogeneity, correlation, shade, and prominence features. The Principal Component Analysis based Discriminant Analysis with Linear Discriminant Analysis (PCA-based DP-LDA) selects the optimal feature selection score values. The Adaboost Support Vector Machine (Adaboost SVM)-based Harris Hawks Optimization (HHO) technique is used to estimate IQ values. The HHO algorithm is used to optimize the number of iterations, error threshold, and weighting factors in AdaBoost SVM, resulting in the estimation of optimal IQ values. Because memory and cognitive abilities are commonly tested in IQ testing, changes in Grey Matter/White Matter (GM/WM) tissues in these ROIs may have an influence on measuring human intelligence. Experimental results show a substantial connection between cortical thickness, laminary difference, and hierarchical position estimation for visual and auditory hierarchies and somatosensory hierarchies. The proposed Adaboost-SVM with the HHO method offers the lowest Root Mean square error value of 8.521 when compared with the existing techniques and gives a classification accuracy of 98% when evaluated with tenfold cross-validation.