The following is an abridged version of the project submitted with the original NSF/NIH proposal. The original project summary is included along with the project description. ? ? Alzheimer's disease (AD) affects an estimated 4 million Americans, making it a major public health concern. While the positive predictive value of clinical diagnosis is around 93% at university research clinics, most patients are evaluated by community healthcare providers where the accuracy of the diagnosis remains uncertain. Meanwhile, active development of pathologically targeted medications requires an accurate diagnosis at the earliest stage possible. ? ? To have a meaningful impact on healthcare, the diagnosis procedure must be inexpensive, non-invasive and available to community physicians. Therefore, this study specially aims: ? (1) to develop and optimize an automated classification system for AD diagnosis that can learn incrementally from data as it becomes available, that can estimate its own confidence as well as the severity of the disease, and that can be implemented in hardware as a low-cost portable device; ? (2) to determine the sensitivity, specificity and positive predictive value of the proposed system in correctly categorizing subjects as demented or normal. ? ? The proposed method combines clinical, mathematical and computational techniques of EEG analysis, multiresolution wavelet analysis and neural network (NN) based ensemble classification, to detect the earliest neurodegenerative changes of AD. While changes in ERPs have been used to identify various neurological disorders, their traditional visual and statistical evaluations have been ineffective in detecting such neurodegenerative changes. Multiresolution wavelet analysis (MWA) represents a promising method for concurrent temporal and spectral analysis of non-stationary ERP signals, whereas NN based ensemble classifiers provide a powerful automated classification tool. The classification system will be optimized to incrementally learn additional information that becomes available over a long period of time. The classifier decisions will then be combined for an automated identification of the subject's clinical classification, as well as the severity of the disease, if the disease is predicted to exist. A statistical analysis of classification results will also be performed to estimate the confidence of the decision. ? ?
| Staudinger, Tyler; Polikar, Robi (2011) Analysis of complexity based EEG features for the diagnosis of Alzheimer's disease. Conf Proc IEEE Eng Med Biol Soc 2011:2033-6 |
| Polikar, Robi; Tilley, Christopher; Hillis, Brendan et al. (2010) Multimodal EEG, MRI and PET data fusion for Alzheimer's disease diagnosis. Conf Proc IEEE Eng Med Biol Soc 2010:6058-61 |
| Ahiskali, Metin; Green, Deborah; Kounios, John et al. (2009) ERP based decision fusion for AD diagnosis across cohorts. Conf Proc IEEE Eng Med Biol Soc 2009:2494-7 |
| Patel, Tejash; Polikar, Robi; Davatzikos, Christos et al. (2008) EEG and MRI data fusion for early diagnosis of Alzheimer's disease. Conf Proc IEEE Eng Med Biol Soc 2008:1757-60 |
| Polikar, Robi; Topalis, Apostolos; Green, Deborah et al. (2007) Comparative multiresolution wavelet analysis of ERP spectral bands using an ensemble of classifiers approach for early diagnosis of Alzheimer's disease. Comput Biol Med 37:542-58 |
| Gandhi, Hardik; Green, Deborah; Kounios, John et al. (2006) Stacked generalization for early diagnosis of Alzheimer's disease. Conf Proc IEEE Eng Med Biol Soc 1:5350-3 |
