General EEG Study Framework

Since 2018, Zentrela has completed numerous IRB-approved studies under Health Canada’s Observational research framework to develop an accurate and objective evaluation of mental-state, mood, and psychoactive effects. These studies employ the use of EEG recordings collected using Zentrela’s end-to-end proprietary Electroencephalography (EEG) system, which aims to capture the distinct electrical neural signatures associated with mental states (e.g. energetic, fatigued etc.). 

An individual’s electrical brain activity is periodically recorded through EEG during a specified trial period, and the complementary physiological and physiological tests are administered in parallel. Depending on the objective of a given study, the individual may ingest a substance (e.g. coffee) or engage in activities (e.g. listening to music) that can induce psychoactivity and shifts in mental states, wherein the resulting EEG recordings and test results thus characterize the expected psychoactive effects. These studies have enabled the accumulation of a substantial EEG database, incorporating many tens of thousands of EEG recordings from over one thousand research participants, and comprises the training and testing data utilized in development of Zentrela’s novel AI solution; the Cognalyzer®. Additional detailed information regarding the specific EEG research protocols is provided in the associated documents for the Z-CUSIC-01, Z-CUSIC-02, and Z-CUSIC-03 studies, as well as the peer-reviewed studies completed by Bosnyak et al., (2022) and McDonald et al., (2021), for reference.

EEG recordings are obtained from patients with Zentrela’s proprietary headset, fitted with 8 non-invasive electrodes strategically placed on right and left sides of the frontal, temporatal, parietal and occipital lobes to collect brain activity from the main regions of the brain. The headset collects and streams data via Bluetooth to a nearby laptop. Data is recorded continuously throughout a given trial using an 8-channel EEG system with a 250-Hz sampling rate and 24-bit resolution. Each recording is meticulously screened for artifacts post-hoc, and unreliable data segments are rejected in real-time. For each measurement, data collection continues until a sufficient amount of reliable, artifact-free segments of EEG recordings are obtained. The relevant psychological and physiological evaluations (i.e. BRUMS, DEQ) are administered concurrently with the EEG measurements. Several measurements are made before (control) and throughout the expected period of mental state alteration or psychoactivity. The resulting data comprises a coarse time-series of detailed brain activity recordings (EEG), mental state/psychoactivity characterizations (BRUMS, DEQ), and physiological measurements (EKG). 

EEG Data Analysis and Labeling

Certain induced and implicit mental states and traits, including fear (Zhou et al., 2021), craving (Garrison et al., 2023), reward (Speer et al., 2023), learning (Kincses et al., 2023 ), and cognition (Sripada et al., 2020), among others, are known to produce distinct neural signatures in aspects such as power spectral density, cross power spectral density, coherence, and RMS power in EEG recordings. Following the EEG recordings and associated data-screening, neurological signatures are extracted from the EEG using Zentrela’s proprietary Cognalyzer® algorithm. The Cognalyzer® scientifically characterizes and labels each EEG recording in regard to the presence/absence and magnitude of certain mental states and emotional traits. Characterization and labeling is completed in accordance with established neural signatures, a proprietary library of neural signatures, and the corroborating psychological and physiological data. 

Sensitivity and specificity of the Cognalyzer® algorithm for cannabis-induced psychoactive mental state predictions have been evaluated and validated by external CRO trials, as detailed in Bosnyak et al., (2022) and McDonald et al., (2021). Zentrela’s unique methodology allows for objective conceptualization of mental states from EEG that has a proven high level of correspondence with subjective evaluations on an individual level. The Cognalyzer® algorithm evolves through continuous refinement, integrating data from numerous individuals to enhance its ability to objectively estimate the psychoactive effect level. Additional labeled data from each individual contributes to refining and validating the algorithm’s performance. Detailed psychographic data collection allows the creation of more discriminators for further enhancing classification. The original labeled EEG data ensures that enhancements can always be reapplied, providing ongoing insights. As such, this methodology provides a generalizable framework to predict and investigate a wider range of mental states and their relation to cannabis, or other substances with perceived psychoactive effects.

Cognalyzer® AI Relaxation Development 

Zentrela has developed a novel AI-based predictive tool designed to identify whether an individual is in a mental state characterized by relaxation. This tool, named the Cognalyzer® AI for Relaxation, was trained and tested using Zentrela’s extensive, proprietary EEG database. Consequently, the Cognalyzer® AI for Relaxation makes predictions based on the EEG recording segments of an individual, whether, and to what extent, a patient is experiencing a mental state associated with relaxation (i.e. sleepiness, calmness, restful etc.). As such, this presents the opportunity to test and evaluate psychoactive substances (e.g. cannabis) in regard to their ability to induce said relaxation effects.

In regards to labeling individual observations with relaxation effect of interest, EEG recording segments were characterized based on established and proprietary neural signatures, and the results of the complementary/non-EEG tools (e.g. BRUMs). The resulting characterizations describe an individual’s mental state in terms of relaxation (calm, de-stressed, hypnagogic) or non-relaxation (energized, alert, hypnopompic), and the magnitude of such. An individual identified as having sufficient indicators of relaxation is coded as “1” (termed “Relaxed”), whereas those with a lack of indicators is coded “0” (termed “Non Relaxed”). 

Consistent with its predecessors (e.g. Cognalyzer® AI for Cannabis’ Psychoactive Effects), the Cognalyzer® AI for Relaxation works in a two-stage process; first processing the EEG data and extracting the relevant features and signatures, and subsequently predicting the mental state of the associated individual. EEG data processing involves signal conversion and extraction of thousands of instances of several EEG features related to this mental state (e.g. frequency range 8-12Hz). Model selection and associated parameter tuning was based on cross-validation on the test set comprising 4155 EEG recordings, wherein the best-performing models were adopted with predictions aggregated via majority-vote. To test performance in practical settings, this solution is then applied to new EEG data collected under controlled scenarios, both before and after the consumption of a product or exposure to external factors that may influence relaxation levels, such as listening to music.