Alzheimer's disease (AD) is considered to be the most common and the fastest growing neurological disease in the world. Biomarker tools for early diagnosis and disease progression in AD remain key issues for clinical applications (patient's control and monitoring), sanitary systems (growth of prevalence of AD in the near future), and pharmaceutical companies (drug developments). Electroencephalogram (EEG) yields, in principle, a powerful and relatively cheap way for screening of dementia and AD in their early stages although not reaching the specificity prescribed for clinical use. Portable EEG systems based on wireless sensors can be used for unobtrusive long term monitoring provided they can solve technological problems (e.g., size and operational lifetime of batteries). Clinical applications require intensive recording of massive EEG data, raising the need for efficient, and flexible compression techniques. In this paper, this compression problem is viewed from a different perspective: is it possible to imagine a compressive sampling/sensing system that not only can reduce the throughput data but can also discriminate among brain states? If this is the case, it would be possible to distinguish among AD patients, mild cognitive impaired (MCI) subjects and normal healthy elderly. Accordingly, this paper investigates the performance of EEG compression techniques to recover data with the quality required by clinical constraints when making an assessment of the patients' status simultaneously. A sparsity coefficient is shown to be a good marker of AD being able to differentiate AD EEG from both MCI and healthy controls. The research, carried out on an experimental database, also compare the thresholding wavelet approach to the emerging method of compressive sensing. The results show the relationships between the compression ratio and the entropic measures of EEG complexity, whose reduction is considered the hallmark of AD
File in questo prodotto:
Non ci sono file associati a questo prodotto.