0424 Visualizing insomnia phenotypes using dimensionality reduction techniques

Abstract Introduction A large number of features can be extracted from a single hypnogram, such as stages durations, onsets, or transitions probabilities. Those numerous indicators turn collection sleep records into high dimension space. Dimensionality reduction techniques are then useful to reveal patterns in data. We used 3 dimensionality visualize insomnia phenotypes dataset and control records: principal component analysis (PCA), t-distributed stochastic neighbor embedding (t-SNE) uniform manifold approximation projection (UMAP). Methods 519 have been included with the following diagnoses: 46 onset insomnia, 83 state misperception, 223 maintenance 117 50 controls (good sleep). limited feature extraction hypnogram macrostructure constitute primary source information for diagnosis polysomnography. common wake after (WASO), well more intricate first set 54 per was computed. were projected 2 dimensional space using PCA, t-SNE UMAP. Results Co-ranking matrix projections computed (PCA: Kmax=103, Qlocal=0.38; tSNE: Kmax=20, Qlocal=0.44; UMAP: Kmax=160, Qlocal=0.44). UMAP technique that hypnograms sets most meaningful way, by reflecting individual diagnoses. Interestingly representation, group outliers, having nevertheless experienced nocturnal awakenings next similar subjects, which indicate fine-grained capture quality at level. Conclusion unsupervised promising methods when approaching spaces. fine clusters could show subgroups already known phenotypes. As literature showed robustness applicability those larger datasets, improved addition new hypnodensities generated machine learning algorithms, spectral polysomnographic signals. Support (if any) None