This is the first study to our knowledge that evaluates permutation entropy in relationship with alteration of awareness in focal seizures. We chose to study a relatively homogeneous group of patients with temporal lobe seizures. Our results reveal a positive correlation between the delta entropy (ΔE) and the consciousness score. Furthermore, we were able to define a threshold of the entropy changes, measured with ΔE, allowing to discriminate between the seizures of group A (no alteration of awareness) and group C (severe alteration of awareness) Another important finding is that in patients with loss of awareness unlike in those with preserved awareness, the demonstrated changes in ΔE values showed more extended spatial distribution.
As far as various disorders of consciousness (DoC, e.g., syncope, coma, vegetative states, sleep disorders, encephalopathy and seizures) are concerned, it is the degree of impairment within the “consciousness system” that defines the extent of alteration of consciousness 28. Numerous studies have been conducted to identify the components of consciousness in patients with DoC and, most importantly, to identify physiological markers in the clinical setting and thereby to better define the patient’s prognosis.
With respect to the structural imaging, findings demonstrated to be specific for patients with permanent DoC (Vegetative state, VS, and minimally conscious state, MCS) as compared to healthy controls, include diffuse abnormalities in the cortical, subcortical, and white matter volume and connectivity 29.
The use of entropy allows a successful evaluation of the "quantity of information" or the "complexity" of a system. Several theories predict that the complexity of information integration 30 or distributed processing 31 is elevated during conscious states. Permutation entropy assesses the regularity of the probabilistic distributions of temporal patterns in the signal 21.
Permutation entropy has been studied by different teams in various DoC. Thul 32 found reduced permutation entropy on scalp EEG in patients with minimally conscious state and in the unresponsive wakefulness syndrome (UWS) compared to healthy controls. Furthermore, this reduction was most pronounced in the UWS group.
Another method developed by King 33 based on weighted symbolic mutual information (wSMI) allowed discrimination between VS and MCS. Sitt and colleagues 34 used different metrics obtained from scalp EEG to discriminate between different states of consciousness (low frequency power, EEG complexity and information exchange). They found that permutation entropy-based measures were particularly efficient in the theta frequency range to discriminate VS from the other groups. PE has been widely used as a tool to track the qualitative effect of anesthetics throughout the different stages (awake, to slightly and deeply sedated) 35, 36 becoming a means of evaluating the depth of anesthesia monitoring 37, 38, 39.
These studies confirmed that access to consciousness requires sufficient information content resulting in high signal complexity whereas altered states of consciousness are associated with low signal complexity.
Li 36 found that permutation entropy can be used to predict absence seizures using genetic absence epilepsy rats from Strasbourg (GAERS) model, and to demonstrate the presence of a pre-ictal state in absence seizures: a dynamic progressive decrease in PE was observed before seizure onset. Furthermore, Ferlazzo 40 showed higher permutation entropy in fronto-temporal areas and lower entropy values in posterior areas, during interictal and ictal periods, in patients with absence seizures in comparison to healthy controls.
The originality of our study is the use of permutation entropy in the context of alteration of consciousness occurring in patients with TLE, recorded with intracerebral electrodes (SEEG). In contrast with other disorders of consciousness, this state is transient and must be captured during the seizure periods. Taken as a whole, we observed a decrease of PE values at the seizure onset and throughout the seizure spread in the brain. This decrease was more pronounced in patients with an alteration of awareness. We estimated the decrease from baseline by a metric called “delta entropy” (ΔE) and we found a positive correlation between the ΔE value and the consciousness score: Group A seizures were characterized by a ΔE value close to zero and Group C seizures with a high negative ΔE. This means that during a seizure with no alteration of consciousness the variation in permutation entropy between the baseline and the minimal entropy value obtained is smaller in comparison to a seizure with severe alteration of consciousness. These results suggest that seizures with maximal alteration of consciousness also have maximal loss of signal complexity. We can thus observe a spectrum of decreased delta entropy as we go from group A to group C, which also applies to the extent of the regions involved as we move along these groups. Indeed, we noted an involvement of numerous brain regions in Group C seizures, while in Group A seizures, mostly the mesial temporal lobe structures (left and right) were involved.
Thus, it appears that the loss of complexity observed in group A seizures is limited to temporal mesial structures. The associative cortical regions keep their level of complexity, probably compatible with information exchange and conscious processing.
In group C seizures, the loss of signal complexity was more diffuse, extending bilaterally as well as to the associative cortices. In term of pathophysiology, these results suggest that the loss of signal complexity becomes incompatible with conscious processing. This is in agreement with different theories of consciousness based on information theory 41, re-entrant processes 42 and global workspace theory13. Noteworthy, no clear pattern was seen in group B seizures, where the delta entropy values were more heterogeneous. This can be explained by the fact that the extent of alteration of consciousness in this group is more diverse, since the consciousness scores vary between 2 and 5 and thus these seizures can behave either like group A or group C seizures.
Furthermore, no positive correlation was found between the delta time (Δtime) and the consciousness score in any group. This could be compatible with the observation that the duration of a seizure does not necessarily determine the extent of alteration of consciousness: an isolated typical epileptic aura can last few minutes and usually does not impair the components of consciousness
Finally, despite some limitations mentioned below, our results suggest that the delta entropy marker can be used as a potential predictor tool for assessing the extent of alteration of consciousness43.