Early Dementia Diagnosis Based on DNN Based Correlational Analysis and Fisher Criterion Based LDA using Morphological Brain Multiplexes

Abstract

The expanding recurrence of dementia happening is a disturbing patterning that has incited dire research intending to avert the improvement of the sickness. Diagnosing dementia in its beginning periods is an urgent advance in averting the improvement of the ailment into exacerbated side effects. Early mild cognitive impairment (EMCI) is the early symptom of dementia. This can be analyzed using mapping mind associations utilizing Magnetic Resonance Imaging (MRI). In the approach, for improving the correlational block, we presented an enhanced classifier also, for improving the performance of discriminative block, an optimized LDA is to be proposed. For correlational analysis, Deep Neural Network (DNN) is presented in this work. Besides, for discriminative analysis, a novel and efficient feature selection method is presented. Fisher criterion is used to select the most discriminatory and appropriate features to ensure consistent feature selection and classifier learning goals and to improve the classifier's performance. In the Mat lab framework this proposed method is implemented. The performance of this proposed approach is evaluated concerning Accuracy, Sensitivity, and Specificity.