Neurodegenerative diseases (NDD) are highly disabling and severe diseases, and become more
common with increasing age. As no cure exist and as the aging population increases, NDDs
are considered to be one of the most serious health problems facing modern society. The
most elusive goal in the field of NDD is to find a neuroprotective agent, and if such treatment
becomes available, it is essential that the patients can be identified as early as possible.
Parkinson’s disease (PD) is the second most common NDD, and early disease identification
is an active field of research as no reliable markers yet exist . Sleep disturbances are
common non-motor symptoms of PD, and strong findings associating a specific sleep disorder
("iRBD") to Parkinsonism suggest that sleep disturbances might precede the clinical diagnosis
of PD. Analysis of sleep thus hold potential to serve as early disease identification, but as the
current standard for sleep analysis relies on manual scorings guided by standards designed to
fit healthy and normal sleep, manual sleep analysis of pathological sleep lacks substance.
This dissertation hypothesizes that automated sleep analysis can identify altered patterns of
EEG and EOG in pathological sleep and may serve to reveal PD biomarkers. The aims of this
dissertation was to: 1) Develop full data-driven sleep models based on EEG, EOG or both, that
can describe sleep in detail and can be used in the analysis of normal as well as pathological
sleep. 2) Extract appropriate features from the automated sleep models describing alterations
in the sleep patterns of patients with PD or iRBD. 3) Identify changes of sleep spindles in the
EEG of patients with PD by extracting features describing spindle morphology.
The results showed that patients with PD or iRBD reflect 1) altered eye movements during
sleep, 2) altered amount and stability of data-determined stages linked to N3 and REM sleep,
3) more REM-NREM sleep transitions determined by a data-driven model, 4) decreased
spindle density and 5) altered spindle morphology compared to non-NDD subjects.
In conclusion, this dissertation illustrates how appropriate biomedical signal processing can
be used to reveal indicative alterations in the sleep EEG and EOG of patients with iRBD and
PD. The automated methods developed analyze sleep in a robust and standardized way and
can be supportive for sleep evaluation. Conclusively, this dissertation contributes to the field
of early PD identification, but substantiates the claim that no known PD biomarker is reliable
enough to stand alone.