Identification of Neuroinflammation and Neuroimaging Biomarkers Through Data Driven Artificial Intelligence Techniques for Unraveling the Heterogeneity of Aged Subjects at Risk of Dementia and to Better Inform Prevention Strategies

Overview

SCD and MCI are very heterogeneous conditions, which can be prodromal to different types of dementia. The application of data driven clustering methods on neuroimaging and inflammatory data aims at identifying the features characterizing subgroups of subjects at high risk of developing overt dementia. This approach promises to develop tailored early interventions for subjects with profiles of cognitive decline correlating with high risk of progression. To this end in this project, we propose to examine in these subjects both brain structure and cerebral blood flow (CBF) through MR T1-weighted and pCASL (pseudo-continuous arterial spin labeling) imaging and the neuroinflammatory status. The latter will be done by measuring the populations of innate and adaptive immune cells, the chemokines that are required to attract these cells to their potential sites of action in the CNS, and the cytokines by means they exert their function. The analyses performed in aged subjects with MCI and in SCD, will allow identifying common mechanisms predisposing to an increased susceptibility to progression to overt cognitive decline.

The recruitment will involve 35 MCI in the training cohort, 35 MCI in the placebo cohort and 75 SCD subjects. Cognitive assessment: A comprehensive battery of neuropsychological tests will assess performance in memory, attention and executive function domains. For all MCI subjects the same cognitive assessment will be performed at the beginning of the study (T0), after 7 months (T7) and after 14 months (T14). SCD subjects will perform the cognitive assessment at T0, T7, T14 and after 28 months (T28), depending on the assigned arm. Instrumental assessment: for all MCI subjects the instrumental assessment, MR brain scans (T1-weighted and pCASL), will be performed at T0 and at T7. 35 SCD subjects will be subjected to the instrumental assessment at T0 and at T28. Blood samples: Blood samples for peripheral blood mononuclear cell (PBMC) isolation will be withdrawn from all MCI and SCD subjects, at baseline. For all MCI patients blood samples will be taken also at T7 and T14, for SCD subjects also at T28. Cellular mediators will be assessed by Flow cytometry analysis of monocyte/macrophage, T cell (CD3/CD4/CD8) and regulatory T cell (CD4+CD25+IL7Rlo). Serum/plasma will be prepared and stored for determination of soluble immune mediators of inflammation. ELISA determinations will be performed for IFNgamma, IL12, TNFalpha, IL-6, CCL2 and BDNF. Intervention: After baseline assessment the MCI-training and SCD-training group will receive a 7 months-long multidomain training, including cognitive, physical exercise and music therapy; while, the MCI-no training and SCD no-training will be followed in their natural history. According to the TtB protocol (Train the Brain Consortium, 2017) enrolled subjects will be assigned to mixed-gender classes of maximum 10 subjects each and given 2 sessions of 60 minutes supervised cognitive training per day, 3 times a week, and one of 60 minutes supervised aerobic physical activity per day, 3 times a week for 7 month. The cognitive training programme will be based on 8 cycles composed by 18 sessions of exercises and activities aimed at stimulating multiple cognitive functions. Each cycle will last 3 weeks, after which the same kind of cognitive stimulation sessions will be restarted, with exercises and activities of increased complexity. Each cycle will be structured in sessions aimed at stimulating specific cognitive domains and in sessions focused on multimodal activities. The physical training will include aerobic exercise on an ergometer cycle, whose duration will be increased gradually from 10 to 20 minutes followed by exercises targeted to improve muscle strength, physical function, neuromuscular control and flexibility. The MCI-training and MCI-no training groups will be compared at T7 in terms of cognitive performances, CBF, gray matter atrophy and inflammatory markers. An additional comparison in terms of cognitive integrity and inflammatory markers will be performed at T14. The SCD-training and SCD-no training groups will be compared at T7 in terms of cognitive performances and inflammatory markers. An additional comparison in terms of cognitive integrity and inflammatory markers will be performed at T14. The MR acquisitions performed on MCI and SCD subjects at baseline (both T1-weighted and pCASL) will be processed in order to obtain measurements on brain structure volumes and CBF. On these measures a consensus clustering technique will be applied in order to identify homogeneous subgroups of anatomical and CBF patterns on both MCI and SCD cohort. First, the MCI and SCD cohorts will be treated separately in order to detect subgroups of the different clinical categories. Second, the MCI and SCD cohorts will be jointly treated in order to detect potential neuroimaging affinity in subjects who exhibit different cognitive impairments. After the first evaluation with consensus clustering approach, novel clustering techniques such as clustering methods based on deep neural network (Xie et al. 2016) will be explored. The robustness of clustering methodologies of neuroimaging data will be tested by a cross validation approach and by enriching the sample with anatomical and perfusion data publicly available thanks to the ADNI consortium (at the moment about 160 anatomical and CBF data are available for a cohort of MCI and SCD subjects).