Poster Design & Printing by Genigraphics - 800.790.4001
Leonard J. Trejo, Ph. D. Roman Rosipal, Ph. D Pacific Development
and Technology, LLC Paul L. Nunez, Ph. D. Cognitive Dissonance, LLC
We developed atomic decomposition algorithms for physiological
estimation of cognitive status, including engagement, cognitive
workload and mental fatigue. We validated the algorithms using EEG
recordings from human participants who performed two different
tasks: Task 1. Simulated control of UAVs (uninhabited aerial
vehicles) Task 2. Continuous mental arithmetic for up to three
hours. Each of two atomic decomposition algorithms identified EEG
atoms with three dimensions: Algorithm 1. EEG power, EEG electrode
position and time on task Algorithm 2. EEG coherence, EEG electrode
pair and time on task. For the UAV control task we found atoms that
combine sources in the theta (4-8 Hz) and alpha (8-12 Hz) EEG
frequency bands consistently at different times of day and on
different days. The temporal variations of the atoms clearly
reflected the levels of mental workload induced by varying task
conditions. For the mental arithmetic task we found atoms that
tracked the development of mental fatigue in individual
participants over time, while reflecting underlying changes in
power or coherence of primarily theta-band EEG. By comparing these
results with results of prior analyses using the same data sets, we
find that atomic decomposition can supplement or overcome existing
approaches based on conventional two- dimensional space-time or
frequency- time decomposition of EEG. Email: [email protected] Web:
[email protected] ADVANCED
PHYSIOLOGICAL ESTIMATION OF COGNITIVE STATUS As exemplified by
Figures 3 and 4, we find that the PARAFAC models allow for robust
estimation of mental workload. The results shown were replicated in
two databases over 14 participants. As shown in Figures 5 and 6, a
PARAFAC model provides for accurate estimation of mental fatigue.
Other results (not shown) confirmed the model in 16 participants.
We also find that inclusion of EEG coherence features in algorithms
for estimating cognitive status improves on algorithms that rely on
the EEG power spectrum alone. Neurocognitive Databases. We used
three databases to develop and test the APECS algorithms: 1.USAF-C:
C2ISR Multimodal Study of UAV Operator Readiness (n=6) 2.USA-T:
Army Multimodal Study of Cognitive Overload (n=8) 3.NASA-C: NASA
Cognitive Fatigue Database (n=16) For each database, we followed a
three-step process of algorithm adaptation and testing (Figure 2).
Prior knowledge of brain functional specialization guided our
selection of EEG electrode sites, frequency bands, and spatial /
temporal resolution for power and coherence. The temporal
resolution of EEG analysis for each task was long enough to
reliably estimate parameters but short enough to detect changes
related to task conditions. By applying these principles to each
task, we produced a set of analysis parameters (Table 1). For tasks
that induced mental workload, we found PARAFAC atoms that combine
sources in the theta and alpha EEG bands consistently in individual
participants at different times of day and on different days. The
temporal variations of the atoms clearly reflected the levels of
mental workload induced by varying task conditions. For a task that
induced mental fatigue, we found PARAFAC atoms that tracked the
development of mental fatigue in individual participants over time,
while reflecting underlying changes in power or coherence of
primarily theta-band EEG. Our results show that APECS algorithms
based on atomic decomposition are valuable new methods for the EEG-
based estimation of cognitive status. By comparing our present
results with results of prior analyses using the same data sets, we
observed that atomic decomposition can supplement or overcome
existing approaches based on conventional two-dimensional
space-time or frequency-time decomposition of EEG. Over the past
four years, we have developed advanced algorithms for real- time
classification of mental states that can run on small hand-held
computers or be embedded in the controls of vehicles, aircraft,
spacecraft, and even in the helmets of foot soldiers, pilots, and
astronauts. The algorithms use methods of machine learning and
statistical pattern recognition, including kernel partial least
squares (KPLS) and parallel factor analyses (PARAFAC). By applying
these methods to massive amounts of experimental data we reduced
multi-sensor EEG spectra to a small set of atoms that are important
for estimating mental states. The algorithms can handle practically
unlimited numbers of input channels and spectral resolutions, are
adaptive, and require no a priori information of the spatial or
spectral distributions of the atoms. Prior research suggests the
hypothesis that cognitive workload is reflected by the
desynchronization of a parietal alpha atom defined by long-range
coherence with frontal regions and the synchronization of a frontal
midline theta atom defined by local coherence with neighboring
frontal regions. Mental fatigue, in contrast, is reflected by
synchronization of both atoms. Our approach was to analyze
multichannel EEG from each task and discover atoms reflecting these
desynchronization and synchronization effects. For this study, we
developed PARAFAC models of EEG power or coherence that allowed us
to discover atoms in three dimensions, { e, f, t }. 1) Power: EEG
power (f ) EEG electrode (e ) Time on task (t ) 2) Coherence: EEG
coherence (f ) EEG electrode-pair (e ) Time on task (t ) Models
were defined by the PARAFAC equation. Each atom is estimated by two
normalized vectors (a,b), a score vector c and a noise term, eft.
Graphical and conceptual illustrations of EEG atomic decomposition
for the power model using PARAFAC are shown in Figure 1.
INTRODUCTION METHODS AND MATERIALS DISCUSSION RESULTS ABSTRACT
Table 1. EEG and Task Analysis Parameters Workload condtions (e.g.,
trials, time) Electrodes EEG Frequency Atoms Molecule with Covalent
Bonds Basic EEG Sources (atoms) EEG atoms (molecule) with Coherence
Bonds Figure 1. Using PARAFAC for EEG atomic decomposition Figure
5. PARAFAC power model during mental arithmetic task performance in
one participant (G). Atom 1 (blue) reflected mostly wide-band noise
from electrode F7. Atom 2 (red) reflected EEG power in alpha and
theta bands. Figure 6. PARAFAC power model during first and last 15
minutes of a 3-hour mental arithmetic task performance in one
participant (G). Atom 1 (blue) did not change over time. Atom 2
(red) reflected the development and progress of mental fatigue.
Estimation of Mental Workload Estimation of Mental Fatigue Figure
3. PARAFAC power model during UAV task performance in one
participant (3). The blue sections reflect periods of low workload
(cruise); red sections reflect high workload (UAV control). Figure
4. PARAFAC coherence model during UAV task performance in one
participant (3). The blue sections reflect periods of low workload
(cruise); red sections reflect high workload (UAV control). Top 10%
of inter-electrode coherences Magnitude squared coherence Copyright
2010 Pacific Development and Technology, LLC Figure 2. Design and
Analysis of APECS Algorithms