Changes in sleep dynamics are increasingly recognized as early biomarkers of neurodegenerative diseases. As part of the George B. Moody PhysioNet Challenge 2026, our team, Biosignal Pilots, developed an algorithmic framework to predict cognitive impairment (CI) from polysomnographic (PSG) recordings. We hypothesize that the interplay between sleep depth and sleep fragmentation, as key components of sleep quality, are able to reveal cognitive decline before clinical symptoms manifest. Our pipeline processes multimodal PSG data, including electroencephalogram (EEG), electrocardiogram, respiratory signals, and clinical annotations, into an extensive feature space. Sleep depth is quantified through slow-wave activity, odds ratio product and spectral band-power ratios, while fragmentation is assessed via sleep event indices, cyclic alternating pattern and related metrices. The feature set is complemented by heart rate variability measures. Different gradient-boosted classifiers (XGBoost, LightGBM, and random forest) were trained on patient-level aggregates. We utilized the Optuna framework for hyperparameter tuning and evaluated performance using stratified 5-fold cross-validation. Shapley additive explanations (SHAP) were employed to interpret feature importance and validate neurophysiological hypotheses. In the unofficial phase of the challenge, our approach achieved a score of 0.724 on the hidden validation set using a XGBoost model. SHAP analysis highlighted the significance of sigma-band metrics, which are associated with sleep depth regulation. In particular, a reduction in mean sigma peak frequency of the EEG was observed, which aligns with established markers of thalamocortical loop degradation and reduced sleep spindles in early stages of neurodegeneration, a risk factor for CI. Furthermore, the temporal distribution of arousals showed high predictive performance, suggesting that sleep fragmentation is a critical marker of brain health. These findings reflect the model's ability to predict future CI diagnoses using sleep depth and sleep fragmentation features, supporting our hypothesis, that sleep quality is a joint product of both sleep depth and fragmentation.