Wavelet-Based Processing and Adaptive Fuzzy Clustering For Automated Long-Term Polysomnography Analysis

To assist in the inspection of sleep-related diagnosis and research, an adaptive method for processing long-term polysomnography (PSG) is proposed in this paper. The extracted features of segmented PSG based on wavelet analysis can be used for clustering the segments with similar pattern into a group. The adaptive fuzzy clustering was used to estimate the clusters within the PSG recordings, the optimal number of clusters and the optimal features of an individual subject. The novel method with the adaptive-to-subject concept exhibits four advantages in comparison with other approaches: 1) full automated, 2) adaptive to the diversity of physiological signals among subjects, 3) less sensitive to noise and artifacts, and 4) effective visualization of analysis results for clinicians. The simulation results show the superiority of the proposed method in long-term PSG analysis