You are here:   Products
  |  Login

Categories

Minimize

All Products

Enquiry List

Minimize

Your Enquiry List is empty

Products

Minimize
b-alert-sft-w
B-Alert Cognitive State GUI
Cognitive States

BIOPAC B-ALERT-SFT-W - B-ALERT COGNITIVE STATE SOFTWARE

B-Alert X10 Analysis Software Add-On; Classify Cognitive States with this analysis software add-on for B-Alert Systems Windows 7/XP OS only. For real-time monitoring of subject fatigue, stress, confusion, engagement and workload, use this B-Alert Cognitive State software with proprietary metrics to classify data from B-Alert Wireless EEG systems.
Code
B-ALERT-SFT-W

BIOPAC B-ALERT-SFT-W - B-ALERT COGNITIVE STATE SOFTWARE

B-Alert X10 Analysis Software Add-On; Classify Cognitive States with this analysis software add-on for B-Alert Systems
Windows 7/XP OS only.

For real-time monitoring of subject fatigue, stress, confusion, engagement and workload, use this B-Alert Cognitive State software with proprietary metrics to classify data from B-Alert Wireless EEG systems. The GUI intuitively represents both the raw and processed data for easy understanding by even the untrained user and up to six systems can run simultaneously on a single Windows 7/XP PC.

To facilitate both real-time and offline analysis, the B-Alert Athena gauges are fully customizable to fit the requirements of the user. In the standard format (shown below), the easy-to-read dashboard gauges (Top Left) and time series (Bottom) windows present B-Alert's highly validated second by second metrics: Engagement, Workload and Drowsiness (along with Heart Rate). Heat maps (Top Right) display EEG power spectral densities (PSD) in both spatial and temporal maps for the traditional Hz bands (Beta, Alpha, Theta, Sigma).



B-Alert Wireless EEG bio-metrics are normalized to an individual subject using 5-minutes of baseline data from three distinct tasks with the sleep onset class predicted from the baseline PSD values. A probability-of-fit is then generated for each of the four classes for each epoch with the sum of the probabilities across the four classes equaling 1.0 (e.g., 0.45 high engagement, 0.30 low engagement, 0.20 distraction and 0.05 sleep onset). Cognitive State for a given second represents the class with the greatest probability. B-Alert cognitive state metrics are derived for each one-second epoch using 1 Hz power spectra densities (PSD) bins from differential sites FzPO and CzPO in a four-class quadratic discriminant function analysis (DFA) that is fitted to the individual’s unique EEG patterns. The table below identifies and briefly describes each baseline task, and associates the task with the B-Alert classification.

Assess Cognitive States from EEG

The B-Alert probabilities should be interpreted in a relative rather than absolute manner. Three standardized baseline tasks normalizes the cognitive state metrics to each individual. High population variability for EEG activity requires individualized model fitting is done for each 1-Hz bin (from 1-40Hz), and is not fit to classic summed bandwidths (i.e. theta, alpha, beta, etc...) to optimize classification measures.

Two individuals will generate somewhat different probabilities for the same task due to a) their innate capability, and b) their state during acquisition of baseline data. If a subject is balancing their checkbook during the eyes closed task, they will not generate as much alpha activity than if they were relaxed. This may increase the occurrence of Distraction probabilities when applied to a different task in which they do mentally relax. Subjects are more aroused the first time they complete a baseline session due to the novelty, so it is preferable to reuse the individual’s DFA to classify new data/sessions. Subjects should avoid consumption of caffeine or nicotine immediately prior to baseline acquisition. Optional acquisition of the extended (20-minute) 3C-VT baseline plus a 7-minute image recognition task allows individuals suffering from fatigue/sleep deprivation or a sleep disorder to be identified.

The B-Alert Workload metric is a generalized model i.e., it is not individually fitted, thus it should also be interpreted in a relative manner. For the linear 2-class workload DFA , probabilities closer to 1 reflect higher workload. EEG workload is correlated with increased working memory load and difficulty level in mental arithmetic and other complex problem solving tasks.

Z-scoring is a useful transformation to convert the relative B-Alert metrics into values that can be compared across subjects or for a repeated-measures within subject experimental design.

  • B-Alert peer-reviewed metrics validation: Johnson, R.R., et al., Drowsiness/alertness algorithm development and validation using synchronized EEG and cognitive performance to individualize a generalized model. Biol. Psychol. (2011), doi:10.1016/j.biopsycho.2011.03.003. Contact us to request a copy.

Output files


The output files generated with the B-Alert software share common formatting features. For example, all file names include a label which described the data. For generated files, one row of data is provided per second (or heart beat) of recording time. The first column lists the subject/session number, the second column the elapsed time (since the start of recording) in hour:minute:second:millisecond (HH:MM:SS:MS), and the third column the system clock time associated with the start of the primary (middle) overlay for the epoch. Output files use a comma separated value (CSV) format for easy import into statistical/analytical software applications. The data file, saved in European Data Format, can be imported directly into AcqKnowledge or, using the script provided in the manual, into MatLab. Click the Specifications tab below for output file details.

Commonly Asked Questions


Q. What scalp sites and EEG frequencies are used for the B-Alert High and Low Engagement, Distraction and Sleep Onset cognitive state metrics?

    A. The four-class DFA is based on bi-polar mid-line recordings at FzPOz and CzPOz using 1-Hz absolute and relative power (1 Hz bin normalized to the sum of the power from 2 – 40 Hz) bins. Of the 12 variables used in the model, one is from the delta/slow theta range (1-4 Hz), two are from the fast theta range (5-7 Hz), five in the alpha range (8-13 Hz), three in the beta range (14-24 Hz) and 12 in the sigma range (25-40 hz).

Q. What scalp sites and EEG frequencies are used for the workload cognitive state metric?

    A. The two-class DFA uses bi-polar recordings from C3C4, CzPOz, F3Cz, F3C4, FzC3, FzPOz using both absolute and relative power. Of the 30 variables, four are from the delta/slow theta range, three from the fast theta range, five from the alpha range, seven from the beta range, and 11 from the sigma range.

Q. If I want to create my own classification algorithm instead of using your cognitive state metrics, what do I need to take into consideration?

    A. The PSD provided by the software are useful inputs to any EEG classifier model. The PSD do not resolve individual differences in EEG amplitude that result from skull thickness, adipose tissue, scalp-electrode impedances, etc. Converting absolute PSD values to relative values assists to some degree but does not overcome all between-subject variability. Z-scoring the PSD will only help to identify the relatively high and low EEG activation across that session for that subject. The best approach is to normalize the PSD to PSD values during a controlled condition. Our research concluded that normalizing to multiple conditions provides the most reliable across-subject classifier. Analyzing the EEG in one-Hz bins provides greater between-subject discrimination than using classic EEG bands (i.e., alpha, beta, etc.). Even with this optimized approach, it does not provide sufficient control over variability to allow the derived metrics to be compared in an absolute manner across subjects.

Q. What are the differences between the B-Alert high engagement and workload gauges?

    A. Measurements of engagement and workload are difficult to define because the terms have fairly broad interpretations and there is no objective gold standard for either. The B-Alert high engagement is typically greater during conditions rich with visual stimuli or during learning. The workload gauge tends to not change very much unless the user is performing concentrated memory encoding/extraction tasks. It is not uncommon for the engagement and workload measures to disassociate. Workload and heart rate variability tends to track more closely. These cognitive state gauges may not exhibit expected changes under all conditions, especially if the level of cognitive effort needed to perform a particular task is underestimated.

Product Documentation

b-alert product sheet
Download
b-alert citations
Download
b-alert operational neuroscience applications
Download