Abstract
Alzheimer's disease (AD), characterized by the presence of neurofibrillary tangles and senile plaques in the brain, is a dementia affecting great part of the elderly population, with an incidence that has increased significantly in the last decades. Thus, early detection of AD becomes a public health issue, as it allows starting treatment able to significantly delay the disease progression. Therefore, development of methods that support early AD diagnosis is of paramount importance. Furthermore, in the search for actual early diagnosis, mild cognitive impairment (MCI) has been shown to be an important risk factor in the development of the disease. Treatment in the prodromal period (MCI) can mitigate the deficits related to disease progression. Clinical diagnostic techniques currently achieve around 90% accuracy in the classification between cases and controls. However, they require the application of a wide range of resources that limit the systematic application of this type of testing to a significant part of the elderly population with supposed memory impairment or subtle cognitive decline. Quantitative electroencephalography (qEEG) is a promising method to overcome this limitation, since it has high availability, low cost and uses relatively simple application procedures. Current automatic EEG-based AD classifiers present accuracy close to 90%. Despite this, the complete automation of the diagnosis through this technique has not yet been achieved. First, signal processing with automatic artifact removal (AAR) techniques is required. Subsequently, it is necessary to extract meaningful features from the EEG signal capable of differentiating between MCI or AD patients and healthy elderly individuals. In this research project we intend to develop, improve and validate biological markers based on the analysis of EEG brain connectivity tools to promote early and accurate diagnosis of both MCI and AD. International patient and control databases will be made available by the partner research group at the Sapienza University of Rome. At the end, integration and optimization of the complete system will be performed, from the AAR to the automatic classification step, in order to reach the whole system's optimum point of operation.
|