Brain states are often described on a single scale from full consciousness to unconsciousness, but this ignores the complex, high-dimensional nature of brain activity. By combining whole-brain modeling, data augmentation, and deep learning, researchers mapped states of consciousness into a low-dimensional space where distances reflect similarities between states. They found an orderly trajectory from wakefulness to brain-injured patients, with coordinates related to functional modularity and structure-function coupling, both increasing as consciousness is lost. Model perturbations provided a geometric interpretation of state stability and reversibility. The work suggests conscious awareness depends on functional patterns encoded as a low-dimensional trajectory within the vast space of brain configurations.
Altered states of consciousness, such as those experienced during dreaming or meditation, offer a way to study how large-scale brain activity relates to different subjective experiences. This paper advocates a research program that combines bottom-up generative models of whole-brain activity, based on known properties of neural tissue, with top-down signatures proposed by theories of consciousness. The authors define altered states, discuss relevant brain-activity signatures, and introduce whole-brain models to explore the mechanisms behind these states. They argue that systematically investigating altered states through bottom-up modeling can clarify the biophysical, informational, and dynamical foundations of consciousness.