P4‐085: An exploratory analysis of variables associated with MCI evolution through data mining with novel artificial neural networks

The need of an early diagnosis of Alzheimer disease prompted the research to distinguish the variables useful to recognize the insidious onset of this pathology. In this study we have tried to assess through a novel type of artificial neural network named Auto Contractive Map (Auto-CM, Semeion) possible cognitive and functional indexes involved in the evolution of MCI. We have shown in previous papers that AutoCM is able to show variables map, preserving their non-linear association, showing connection schemes and capturing the complex dynamics of adaptive interactions. From 781 subjects with MCI diagnosis (Petersen, et al. 1999), 683 were excluded for the following reasons (MMSE<23, other neurological and psychiatric diseases found during the interview or in pharmacological treatment, liver or thyroid diseases, missing data). Ninety-four MCI subjects, with clinical and neuropsychological evaluation and 1 year follow-up were included in the study. The total number of variables, (n=86), were obtained in reference of equivalent scores methodology (Spinnler, et al. 1987; Capitani, et al. 1990). The Auto-CM system showed a series of non linear associations between prodromal condition of AD and low scores on immediate and delayed recall of verbal memory test (15 words of Rey), as well as delayed recall of visuo-spatial material (Figure of Rey). The MCI who progressed to dementia had also deficit on Stroop Test (Errors and Time) and a reduced level of functional activities of daily living (BADL and IADL). The preservation of these variables was found to be not associated to the progression of MCI, while the cognitive normalization at follow-up seemed to be linked to highest cognitive indexes of verbal, visuo-spatial memory and extra-mnestic domains. This study provided evidence that the verbal and visuo-spatial memory, executive functions and the preservations of activities of daily life might play a key role in predicting the evolution of MCI subjects. The Auto-CM system could be an useful and novel approach to conduct research in a framework that is characterized by a redundancy and complexity of variables interplay. Future studies including MCI with extra-mnestic deficits could help to better define this cognitive syndrome.