Intelligent quotient estimation from MRI images using optimal light gradient boosting machine

Abstract

Due to the complex structure of human brain, the intelligence quotient (IQ) score estimation becomes a complicated task. The brain variables such as brain volume, gray matter and white matter are significantly contributing toward the measure of human intelligence. In this paper, a novel rectified linear unit-bidirectional long short-term memory with light gradient boosting machine-based dynamic membrane-driven bat algorithm (RBLGBM-DMB) technique is proposed to estimate IQ values using magnetic resonance imaging (MRI) dataset. In order to measure IQ score, the proposed model incorporates diverse steps such as pre-processing, segmentation feature extraction and estimation phases. In this, the pre-processing steps use some common pipelines to enhance data quality along with skull stripping and slice extraction processes to increase model effectiveness in estimating IQ value. Here, MeVisLab is utilized for segmenting the region of interest from MRI images. The ReLu-BiLSTM with light GBM model efficiently extracts detailed information from the input data. However, the over-fitting problems in the light GBM model may have the possibility to influence the proposed method’s performance. Therefore, the DBM algorithm is incorporated with the classifier to minimize the over-fitting problem. To validate the effectiveness of this proposed method, the simulation is performed using MATLAB software with tenfold cross-validation. The proposed RBLGBM-DMB technique offers higher classification accuracy of about 97.8%, and the RMSE rate with and without age matrix is 0.12 and 0.9. It reveals that the proposed technique achieves greater results in estimating IQ scores than other state-of-the-art techniques.