RaPID-2 : Resources and ProcessIng of Linguistic , Para-Linguistic and Extra-Linguistic Data from People with Various Forms of Cognitive / Psychiatric Impairments

In the past decade the preclinical stage of Alzheimer’s Disease (AD) has become a major research focus. Subjective cognitive decline (SCD) is gaining attention as an important risk factor of AD-pathology in early stages of mild-cognitive impairment (MCI), preclinical AD and depression. In this context, neuropsychological assessments aim at detecting sorts of subtle cognitive decline. Automatic classification may help increasing the expressiveness of such assessments by selecting high-risk subjects in research settings. In this paper, we explore the use of neuropsychological data and interview based data designed to detect AD-related SCD in different clinical samples to classify patients through the implementation of machine learning algorithms. The aim is to explore the classificatory expressiveness of features derived from this data. To this end, we experiment with a sample of 23 memory-clinic patients, 21 depressive patients and 21 healthy-older controls. We use several classifiers, including SVMs and neural networks, to classify these patients using the above mentioned data. We reach a successful classification based on neuropsychological data as well as on cognitive complaint categories. Our analysis indicates that a combination of these data should be preferred for classification, as we achieve an F-score above 90% in this case. We show that automatic classification using machine learning is a powerful approach that can be used to improve neuropsychological assessment.