Calibrated voxel data from fMRI for the detection of developmental dyslexia using hybrid CNN-SVM model

Dyslexia, one of the most prevalent learning disabilities among children, is a neurological syndrome that impacts children's literary proficiencies.In the present study, we calibrate the complex voxel information present in functional magnetic resonance imaging using normalising, reducing and flattening techniques, employ dimensionality reduction over the flattened voxel data using principal component analysis and linear discriminant analysis, and then apply artificial neural network and support vector machine (SVM) for classifying into dyslexic and non-dyslexic.In the second methodology, we directly hand over the normalised voxel data to our hybrid convolutional neural network (CNN) and SVM model and acquire an accuracy of 94.44% an average precision of 95%, an average recall of 94.5%, and an F1-score of 94.5%.The CNN-SVM model continues to perform robustly when validated on an independent dataset.