Automated Fiber Quantification for Alzheimer’s Disease Diagnosis from Neuropsychological Biomarkers

Alzheimer’s disease is the gravest among the neurodegenerative diseases because of its high prevalence and large mortality rate. Its manual revelation has become clinically insignificant because of the in expertise, high rate of false positives (FP), and false-negative (FN). This paper frames a quick, affordable, and objective judgement of AD with a novel data mining method. The paper utilises the upgraded flavour of support vector machine-assisted with quantile hyper-spheres (QHSVM) for cataloguing, which frames quantile hyper-circle by utilising pinball loss in place of hinge loss, and thus, makes the model insensitive towards the noise. The trial results on the benchmark ADNI dataset displayed the QHSVM when equated with SVM, Pin-SVM, TWSVM, and THSVM through experimental investigations, which justified the better performance of the QHSVM, even on the addition of noise. The results revealed that SVM got an accuracy of 96.87%, which got dipped to 94.50% on the incorporation of noise. The QHSVM got 97.89% as the initial accuracy and 97.65% as the accuracy value after incorporation of noise, and thus, justifies the least effect of added noise on QHSVM.