AI-Mind Project for Dementia Prediction with time series, genetics, blood markers, cognitive and medical data (1.0)

The AI-Mind is an ongoing project that contains longitudinal multimodal data from approximately 1040 participants recorded at four time points over the course of two years. The project will use the data to develop and train two artificial intelligence-based models to give a risk estimate for developing dementia in people with mild cognitive impairment (MCI). The AI-Mind project uses non-invasive functional imaging techniques: 126-channel EEG (ANT Neuro) and 306-channel MEG (MEGIN). In addition, blood is drawn for biomarker analysis of p-tau181 and 217 and APOE-allele sequencing. The data also contain cognitive and behavioural scores using the CANTAB™ assessments, and socio-demographic and medical questionnaires. Automatic processing steps for source reconstruction of the electrophysiological data is based on a forward modelling method using an anatomically realistic boundary element model (BEM), combined with linearly constrained minimum variance (LCMV) and dynamic imaging of coherent sources (DICS). In addition, a brain-template-based forward models coupled with linear inverse modelling via exact low-resolution brain electromagnetic tomography (eLORETA) is used. Feature extraction is based on both manual and automatic methods (e.g., corrected imaginary phase locking value, ciPLV, and Bayesian reduced rank regression, BRRR). Manual feature extraction is guided by a priori information, reducing the necessity for multiple comparison correction and the type-II error rate (false negatives). For more information, please visit [https://cordis.europa.eu/project/id/964220/results](https://cordis.europa.eu/project/id/964220/results).