Multimodal spatio-temporal-spectral fusion for deep learning applications in physiological time series processing: A case study in monitoring the depth of anesthesia

•Fusing multi-modal physiological data based on time-frequency ridge.•Compressing physiological data directly at the source.•Embedding spatial, temporal, spectral characteristic in just a unified representation.•Transfer learning from image-based architectures to timeseries classification.•Converting physiological timeseries classification into a visual pattern recognition.