Consciousness, and the ways in which it can become impaired after certain brain injuries, are not well understood, making disorders of consciousness (DOC), like coma, vegetative states and minimally conscious states difficult to treat. But a new study, published in Nature Neuroscience, indicates that AI might be able to help researchers gain some traction with this problem. The research team involved in the new study has developed an adversarial AI framework to help them determine what exactly is going on in states of reduced consciousness and how to approach a solution.
Two AI models play a consciousness game
To better understand the mechanisms behind impaired consciousness, the researchers developed two types of AI models and had them play a kind of game where one model determined different levels of consciousness based on EEGs simulated to look like those of real unconscious and conscious brains. The AI agents guessing consciousness levels, called deep convolutional neural networks (DCNNs), were first trained on 680,000 ten-second recordings of brain activity from conscious and unconscious humans, monkeys, bats and rats to detect which neural signals related to differing levels of consciousness. The AI showing EEG data was a biologically plausible simulation of the human brain.
“To decode consciousness from these signals, we trained three separate DCNNs, each specialized for a different brain region, to output a continuous score from 0 (unconscious) to 1 (fully conscious): a cortical consciousness detector (ctx-DCNN), a thalamic consciousness detector (th-DCNN) and a pallidal consciousness detector (pal-DCNN). The ctx-DCNN was trained on continuous consciousness levels derived from clinical scales (GCS and CRS-R), enabling it to recognize graded states of consciousness,” the study authors explain.
Without explicit programming, the AI model was able to deduce known responses to brain stimulation that occur in DOC. The team then analyzed the parameters that the simulation model tweaked in order to find testable predictions about the underlying mechanisms of unconsciousness.
Revealing new mechanisms behind unconsciousness
The researchers say that the model predicted two previously unknown mechanisms for unconsciousness that they were able to validate. The first is an increased inhibitory-to-inhibitory neuron coupling in the cortex, in which more neurons are restraining the firing of other neurons. This results in reduced overall activity. The researchers were able to validate this prediction from RNA sequencing data of brain tissue from comatose patients and in data from rats with brain damage from strokes. The team found that those with impaired consciousness showed an upregulation of genes that drive cortical inhibitory synapse formation.
The AI model also predicted that those with impaired consciousness have a selective disruption of the basal ganglia indirect pathway—a neural circuit that increases inhibition of the thalamus, thereby suppressing unwanted movements and motor actions. To validate the prediction, the researchers analyzed diffusion tensor imaging (DTI) scans from 51 patients with different DOC disorders. They say their analysis provided supporting evidence for the plausibility of selective basal ganglia pathway disruption in pathological unconsciousness, although some limitations, like a lack of cell-type specificity in DTI, of the study warrant further validation studies.
A new target for ‘waking up’ the brain
Although deep-brain stimulation (DBS) has shown promise for DOC therapy in previous studies, the method suffered from a lack of clear mechanistic targets. But the AI model in this study seems to have identified high-frequency stimulation of the subthalamic nucleus (STN), in particular, as a promising target. And the team says there is good indication that this is a reasonable target. They cite a previous study on people with an implanted DBS device for a type of neck spasm, in which some patients had stimulation to the subthalamic nucleus. However, the patients in that study were conscious, so more work is needed to test out the theory.
“Critically, our framework offers a platform for in silico testing of DBS strategies in DOC. Across modeled cases, high-frequency (50–130 Hz) stimulation of the STN consistently increased the AI-predicted level of consciousness. In patients with cervical dystonia, who were fully conscious during both the DBS on and DBS off conditions, predicted consciousness levels were already nearly maximal but showed a consistent upward shift with stimulation. This indicates that STN DBS may push neural dynamics toward patterns classified as more conscious-like by the DCNN, rather than restoring lost consciousness in this context,” the study authors write.
The team hopes to refine their work in future studies and potentially test out whether high-frequency stimulation of the subthalamic nucleus can actually “wake up” patients with impaired consciousness. As for the adversarial AI framework, they say similar methods may be adapted for other complex brain disorders.
A Research Briefing on the work was also published in Nature Neuroscience.