干旱区生态环境知识资源中心

The Alzheimer's Disease (AD) is a disorder neurodegenerative, which is one of the most common causes of dementia in the older people, it constitutes one of the diseases with great social impact in Europe and America. The progress of medical diagnosis using Magnetic Resonance Imaging (MRI) is widely used for the treatment of neurological diseases; it allows obtaining, increasingly, more functional and anatomical information from the brain of the patients, with a great precision in time and space. However, this big amount of data and images are impossible to analyze directly, is necessary to develop methodologies of calculus for quantified the parameters more relevant in the MRI.

This work proposes a methodology for the diagnosis of dementia based on Alzheimer's disease combining imaging processing and artificial intelligence techniques. We created an Artificial Neural Network (ANN) of classification based on architecture Multilayer Perceptron. In order to construct a complete dataset for training and testing of the network, initial inputs-target variables were obtained from the database OASIS (Open Access Series of imaging studies) with a total of instances equal to 235. The variables were classified into 3 groups: demographic, clinical and morphometric data. Task of training and testing were applied on initial data, obtained a 48% of confusion of the diagnosis. For minimize this percentages of error, image processing and Voxel Based Morphometry (VBM) techniques were implemented to obtain new morphometric variables of three areas of the brain: White Matter (WM), Gray Matter (GM) and fluid (CSF) cerebro-espinal. In this way, we reduced the percentage of confusion to 17%. The results obtained with the ANN, demonstrated that the demographic and clinical information from patients, combined with morphometric information of areas of the brain, are input variables useful to train an ANN of diagnosis of dementia with 83% of reliability, and in this way, help to the early diagnosis of AD.