Assessing Significance of Cognitive Assessments for Diagnosing Alzheimer's Disease with Fuzzy-Rough Feature Selection

Research in dementia diagnosis typically involves a range of data modalities and also, the use of cognitive assessments, aiming at the development of approaches that are non-invasive, time-saving and economical. Given the existing diversity of prevalent cognitive assessment factors it is useful to assess and exploit the effectiveness of such cognitive features, while working towards the establishment of a methodology for making informed choice of such factors in practical use. As an initial approach, this paper employs the powerful Fuzzy-Rough Feature Selection (FRFS) technique to support such an analysis, by varying the underlying similarity functions and search strategies employed by FRFS. Evaluated on a benchmark from the renowned Alzheimer’s Disease Neuroimaging Initiative repository, experimental results demonstrate the significance and predictive capabilities of different cognitive assessments in working with a variety of popular classifiers.