AUTOMATED INTELLIGENT DIAGNOSTIC OF ALZHEIMER DISEASE BASED ON NEURO-FUZZY SYSTEM AND DISCRETE WAVELET TRANSFORM

Abstract

Alzheimer's disease (AD) is the deterioration of cognitive functions such as problem-solving, memory and reasoning that interferes with an individual's daily functioning. Symptoms are exhibited in accordance to the affected brain area (e.g. language and learning). Magnetic Resonance Imaging is the best diagnostic method to monitor the damage in brain tissues. Therefore, this paper proposed a hybrid method to segregate between the MRIs of healthy subjects and those diagnosed with AD. In this combined method, around 70 MRIs collected from local hospitals of Jordan are analyzed based on the use of Low-pass morphological filters and equalizer filter. Discrete wavelet transform (DWT) described as mathematical functions that discriminate data into different frequency components, and then each component is considered as resolution — scale matching. In this work, we used the best DWT type namely Haar because it is satisfactory for most of applications of image processing. The DWT then applied on the same MRIs where the entropies and energy values were extracted from each sub-band of the third-level wavelet coefficients to be fed into adaptive neuro-fuzzy inference system (ANFIS) classifier. The ANFIS was trained with the back-propagation gradient descent method in combination with least squares method. The combined method has demonstrated the best classification accuracy of 93% which can be reliably used for diagnosis purposes.