Jack Silcox and Bruce Brown, Department of Psychology
Introduction
Alzheimer’s disease is a progressive neurodegenerative disease that always results in death. Unfortunately, the death that eventually comes is preceded by vicious symptoms. A patient with the disease slowly begins to lose memory and other cognitive abilities, robbing a person of their mind and life.
Alzheimer’s disease is becoming a major public health problem (Blennow, de Leon, & Zetterberg, 2006). In 2006, it was estimated that about one quarter of people of above the age of 85 years suffer from Alzheimer’s disease (Ferri, Prince, Brayne, Brodaty, Fratiglioni, Ganguli,… Scazufca). Because of how destructive this disease is to individuals, families and the community, the diagnosis and treatment of this disease is paramount. Unfortunately, current means of diagnosis are not sufficient. Techniques rely on diagnosing behavioral symptoms that have arisen in a patient (McKhann, Knopman, Chertkow, Hyman, Jack Jr., Kawas,… Phelps, 2011). But, if the behavioral symptoms have begun to manifest, this means that neurodegeneration has already occurred and there is no way to reverse this. Therefore, there are many groups that are working on ways to diagnose the disease early, ranging from biochemical assays to neuroimaging techniques. In this study we looked at the predictive capacity of data collected from an electroencephalogram (EEG). If the EEG can be used in this diagnostic way then it would provide a nonintrusive, inexpensive way to diagnose Alzheimer’s disease.
Methodology
Event-related potential data was recorded from 75 subjects that met NINCDS-ARDA criteria for mild Alzheimer’s disease. 95 healthy controls were also recorded for comparison. Data were collected from an EEG cap that had seven electrodes (Fz, Cz, Pz, F3, P3, F4, and P4). Subjects were asked to perform an auditory recognition task where they were instructed to identify a “target” tone that was shown to them. They were then presented with various tones, one at a time, and asked to identify whether or not it matched their target.
Previously, we have shown that an eigenvector-based method could differentiate between the mild Alzheimer’s group and the controls (Hendrix, et al., 2015). This was done by performing a principle component analysis on one subject and one electrode location at a time. The principal components were then regressed to extract cognitive components that corresponded with the various experimental conditions. These “cognitive components” were then used to differentiate between the two groups. With this project we wanted to test the predictive power of these cognitive components. To do this we used the cognitive components as the predictor variables in a logistic regression. We then found the sensitivity and specificity as measures of how well these components would do in a diagnostic setting.
Results
After determining which model provided the best predictive power a cross validation study was performed using the leave-one-out method. Sensitivity was calculated to be 0.413. Specificity was calculated to be 0.611. Misclassification rate was calculated to be 0.587.
Discussion and Conclusion
These results were surprising considering the high F-ratios we found in differentiating the two groups last year (Hendrix, et. Al, 2015). While it appears that this data shows that there are measurable differences in the task related potentials, they do not do the best job at predicting which of the two groups a patient belongs to.
The failure to have high rates of prediction may have to do with our methodology and we are working at this time to improve our technique. It could also be because the task is not ideal for a clinical setting where diagnosis is important. Rahayel and colleagues (2012) have found that Alzheimer’s patients have major deficits in their smell sensations and perceptions that can be measured behaviorally. Perhaps a future study could be done with a smelling task instead of an auditory task like the present study has done.
Finding the appropriate task while being measured by an EEG would be the next step in this line of questioning. Our past results and our current results suggest that although this task is not ideal there was a measurable difference detected. If we can zero in on the best task to use, the EEG could be a cheap and noninvasive way to diagnose Alzheimer’s disease.
References
Blennow, K., de Leon, M.J., & Zetterberg, H. (2006). Alzheimer’s disease. Lancet, 368(9533), 387-403. doi:10.1016/S0140-6736(06)69113- 7
Ferri, C. P., Prince, M., Brayne, C., Brodaty, H., Fratiglioni, L., Ganguli, M., . . . Scazufca, M.Global prevalence of dementia: A delphi consensus study. The Lancet, 366(9503), 2112-2117. doi:http://dx.doi.org/10.1016/S0140-6736(05)67889- 0
McKhann, G. M., Knopman, D. S., Chertkow, H., Hyman, B. T., Jack Jr., C. R., Kawas, C. H., . . . Phelps, C. H. (2011). The diagnosis of dementia due to Alzheimer’s disease: Recommendations from the national institute on aging-Alzheimer’s association workgroups on diagnostic guidelines for alzheimer’s disease. Alzheimer’s & Dementia, 7(3), 263-269. doi:http://dx.doi.org/10.1016/j.jalz.2011.03.005
Brown, B.L., Hendrix, S.B., Cecchi, M., Scott, J.M., Silcox, J.W., Brighton, K.D., and Hedges, D. (2015): A Novel EigenvectorBased Method to Detect Mild Alzheimer’s Disease Using EventRelated Potentials. The Journal of Prevention of Alzheimer’s Disease (JPAD). http://dx.doi.org/10.14283/jpad.2015.79