In this study comparative data on sleep and its features (macro- and microstructure, presence of sleep disorders) of cognitive impaired patients (predementia/dementia forms) and age and sex-matched healthy controls, are presented; the overall aim is to assess possible biomarkers of early diagnosis and progression of dementia. Initially, 30 amnestic mild cognitive impairment – aMCI patients (19 males, 11 females mean age 72.9 6.6 years) vs. 18 mild-moderate Alzheimer’s disease – AD patients (9 males, 9 females m.a. 70 10.6) vs. 24 healthy elderly (14 males, 10 females m.a. 72.7 12.7) are compared. Both macro- and microstructure data, show a “continuum” regarding sleep features deterioration along the cognitive impairment (e.g., increase of stage 1, decrease of stage2 and NREM/REM cycles, increase of apnea severity, loss of sleep spindle –SS– coherence between the hemispheres). To better assess sleep microstructure, an automatic system for SS detection in demented and pre-demented forms was developed, by using 2 AD patients (1 male 75 ys, 1 female 76 ys), 7 aMCI (7 females m.a 75.7 5.8) and 2 different control groups: A (3 males, 26 ys), B (3 males, 1 female m.a. 67 7). The reliability of this Artificial Neural Network Automatic SS Detection System – ANNASSDS – was proven. Then data on a comparative study between early AD vs early Frontotemporal Dementia (behavioral variant) – bvFTD patients are presented. 12 AD (7 males, 5 females, m.a. 64.5 ± 8.0) vs 12 FTD patients (7 males, 5 females, m.a. 61.0 ± 7.0) of comparable age, cognitive impairment, but with significantly less disease duration for FTD, are compared. The overall sleep profile (e.g. sleep stages latencies, NREM/REM cycles) is worse for FTD. Given that primary sleep disorders (e.g. sleep apnea) equally burden both groups, in order to explain these results a “central” etiology theory is proposed. The ANNASSDS is implemented to investigate if characteristics of SS, which have a protective role in sleep, and their distribution, could provide an insight on the above-mentioned “central” theory. The results are interesting: SS in both groups are equally decreased compared to normative values; SS decrease in FTD is shown for the first time. Moreover SS distribution matches with the pathological lesions; hence, SS in frontal regions are more prominent among AD patients while they are decreased in FTD.