An Efficient Hybrid Deep Neural Network for Intelligence Quotient Detection from Brain Magnetic Resonance Imaging

Accurate IQ detection through non-invasive methods, such as brain MRI analysis, can play a critical role in the early diagnosis and prevention of neurological disorders. This study proposes a hybrid approach that leverages the feature extraction power of Convolutional Neural Networks (CNN) and the robust classification capabilities of the XGBoost algorithm to enhance the accuracy of IQ prediction models. CNNs effectively process MRI images by extracting complex visual features, while XGBoost optimizes classification performance through ensemble learning techniques. The integration of these methods results in a more precise and reliable IQ detection system. Furthermore, the model's architecture and data processing techniques offer potential for adaptation in web-based diagnostic platforms and large-scale, distributed medical research applications. This study presents an innovative approach by combining convolutional neural networks (VGG16 and ResNet-50) with the XGBoost algorithm to classify intelligence levels from brain MRI images. Unlike previous studies that used CNNs solely for feature extraction, this research employs XGBoost for final classification, achieving an 83 % accuracy in the VGG16+XGBoost model, showing a notable improvement over the 73 % reported in the most similar prior study. The use of 3D NIfTI data and sagittal slices for enhanced data analysis is also a key innovation of this research.