Selected Validation Across Tasks, Participants, and EEG Configurations
InnoBrain’s Human Cognitive Monitoring platform has been evaluated across multiple independent datasets, cognitive tasks, EEG devices, and recording environments.
Our validation studies include:
- Controlled laboratory experiments
- High-fidelity driving simulators
- Full-scalp, reduced-channel, and in-ear EEG
- Multiple EEG hardware configurations
- Previously unseen participants and tasks
Links to the complete studies and publications are included in each section.
EEG Signal Cleaning
InnoBrain’s proprietary cleaning algorithm detects and reduces ocular, muscular, cardiac, movement-related, and environmental artifacts before cognitive-state analysis.
Validated against expert manual annotations
| Metric | Result |
| Recall | 91% |
| Precision | 81% |
The algorithm has been evaluated across laboratory, simulator, and wearable EEG recordings.
Explore the cleaning algorithm validation
Mental Workload: CORTEX™
The CORTEX™ model series continuously estimates changes in mental workload from EEG.
CORTEX-O
The model was trained using a controlled arithmetic dataset and evaluated through Leave-One-Subject-Out Cross-Validation.
Laboratory results
- 20 participants
- 21-channel dry EEG
- 75.4% classification accuracy
Cross-Task Driving Simulator Validation
To demonstrate cross-task generalization, the same model was applied, without driving-specific retraining, to 64 simulator recordings from 32 previously unseen participants across underload and overload driving conditions using a different 8-channel EEG configuration.
The model detected significantly higher workload during the more demanding driving condition, consistent with NASA-TLX assessments.
This evaluation demonstrates generalization across participants, tasks, and EEG configurations.
Explore the CORTEX-O validation
CORTEX-O2
CORTEX-O2 was evaluated using an arithmetic dataset involving 20 participants.
| Metric | Result |
|---|---|
| ROC AUC | 0.81 |
| F1 score | 0.85 |
| Balanced accuracy | 0.82 |
The model also showed a graded workload response as arithmetic difficulty increased:
Easy: 0.39 → Medium: 0.53 → Hard: 0.69
Explore the arithmetic task validation
Cross-Task Stroop Validation
CORTEX-O2 was additionally applied to a separate Stroop dataset involving 20 participants, without Stroop-specific training.
The model identified the highest proportion of high-workload EEG windows during the learning of new task rules and the 50% incongruent condition phases. These findings demonstrate sensitivity to both rule acquisition and cognitive interference in a previously unseen task.
Explore the Stroop validation
Fatigue Spectrum™: DRIFT-O™
DRIFT-O™ monitors the neurological progression across the Fatigue Spectrum™:
Alert → Fatigued → Drowsy → Sleep Onset
The model has been evaluated using scalp EEG, reduced-channel simulator recordings, and wearable in-ear EEG.
| Validation Setting | Evaluation Data | ROC AUC | F1 score | Balanced accuracy |
|---|---|---|---|---|
| Scalp EEG, Laboratory PVT | 23 participants | 0.90 | 0.87 | 0.88 |
| Scalp EEG, cross-task driving simulator validation | 102 datasets from 34 participants | 0.88 | 0.74 | 0.76 |
| In-ear EEG, Laboratory PVT | 18 participants | 0.93 | 0.95 | 0.94 |
The cross-task driving simulator validation included three driving environments:
- Highway
- Country roads
- City driving
It also included varying EEG devices, channel configurations, and higher operational noise than controlled laboratory studies.
The in-ear study demonstrated that DRIFT-O can detect fatigue-related neural changes using a minimally intrusive wearable EEG configuration.
Explore the scalp EEG validation
Explore the simulator validation
Explore the in-ear EEG validation
Designed for Generalization
Across these studies, InnoBrain models have been evaluated for:
- Subject generalization: Testing on previously unseen participants
- Task generalization: Training on one task and applying the model to another
- Hardware robustness: Evaluation across different EEG devices and channel configurations
- Environmental robustness: Testing across laboratories, simulators, and wearable settings
Additional Metrics
InnoBrain’s broader monitoring platform also includes:
Calmness · Situational Awareness · Stress · Concentration · Motion Sickness
These metrics are grounded in established EEG and cognitive-neuroscience research, with additional validation studies ongoing.
| Model | Robustness & Generalization | Recording Configurations | ||
|---|---|---|---|---|
| Cross-Task | Subject-Independent | Scalp EEG | In-Ear EEG | |
| CORTEX-O | ✓ | ✓ | ✓ | — |
| CORTEX-O2 | ✓ | ✓ | ✓ | — |
| DRIFT-O | ✓ | ✓ | ✓ | ✓ |
✓ Validated — Not currently demonstrated