Kevin Butz

Comparison of EEG-features and classification methods for motor imagery in patients with disorders of consciousness.

Current research aims at identifying voluntary brain activation in patients who are behaviorally diagnosed as being unconscious, but are able to perform commands by modulating their brain activity patterns. This involves machine learning techniques and feature extraction methods such as applied in brain computer interfaces. In this study, we try to answer the question if features/classification methods which show advantages in healthy participants are also accurate when applied to data of patients with disorders of consciousness. A sample of healthy participants (N = 22), patients in a minimally conscious state (MCS; N = 5), and with unresponsive wakefulness syndrome (UWS; N = 9) was examined with a motor imagery task which involved imagery of moving both hands and an instruction to hold both hands firm. We extracted a set of 20 features from the electroencephalogram and used linear discriminant analysis, k-nearest neighbor classification, and support vector machines (SVM) as classification methods. In healthy participants, the best classification accuracies were seen with coherences (mean = .79; range = .53−.94) and power spectra (mean = .69; range = .40−.85). The coherence patterns in healthy participants did not match the expectation of central modulated -rhythm. Instead, coherence involved mainly frontal regions. In healthy participants, the best classification tool was SVM. Five patients had at least one feature-classifier outcome with p0.05 (none of which were coherence or power spectra), though none remained significant after false-discovery rate correction for multiple comparisons. The present work suggests the use of coherences in patients with disorders of consciousness because they show high reliability among healthy subjects and patient groups. However, feature extraction and classification is a challenging task in unresponsive patients because there is no ground truth to validate the results.

Connectivity biomarkers can differentiate patients with different levels of consciousness

OBJECTIVE In the present study, we searched for resting-EEG biomarkers that distinguish different levels of consciousness on a single subject level with an accuracy that is significantly above chance. METHODS We assessed 44 biomarkers extracted from the resting EEG with respect to their discriminative value between groups of minimally conscious (MCS, N=22) patients, vegetative state patients (VS, N=27), and - for a proof of concept - healthy participants (N=23). We applied classification with support vector machines. RESULTS Partial coherence, directed transfer function, and generalized partial directed coherence yielded accuracies that were significantly above chance for the group distinction of MCS vs. VS (.88, .80, and .78, respectively), as well as healthy participants vs. MCS (.96, .87, and .93, respectively) and VS (.98, .84, and .96, respectively) patients. CONCLUSIONS The concept of connectivity is crucial for determining the level of consciousness, supporting the view that assessing brain networks in the resting state is the golden way to examine brain functions such as consciousness. SIGNIFICANCE The present results directly show that it is possible to distinguish patients with different levels of consciousness on the basis of resting-state EEG.

Real movement vs. motor imagery in healthy subjects

Motor imagery tasks are well established procedures in brain computer interfaces, but are also used in the assessment of patients with disorders of consciousness. For testing awareness in unresponsive patients it is necessary to know the natural variance of brain responses to motor imagery in healthy subjects. We examined 22 healthy subjects using EEG in three conditions: movement of both hands, imagery of the same movement, and an instruction to hold both hands still. Single-subject non-parametric statistics were applied to the fast-Fourier transformed data. Most effects were found in the α- and β-frequency ranges over central electrodes, that is, in the μ-rhythm. We found significant power changes in 18 subjects during movement and in 11 subjects during motor imagery. In 8 subjects these changes were consistent over both conditions. The significant power changes during movement were a decrease of μ-rhythm. There were 2 subjects with an increase and 9 subjects with a decrease of μ-rhythm during imagery. α and β are the most responsive frequency ranges, but there is a minor number of subjects who show a synchronization instead of the more common desynchronization during motor imagery. A (de)synchronization of μ-rhythm can be considered to be a normal response.

EEG-Response Consistency across Subjects in an Active Oddball Task

The active oddball paradigm is a candidate task for voluntary brain activation. Previous research has focused on group effects, and has largely overlooked the potential problem of interindividual differences. Interindividual variance causes problems with the interpretation of group-level results. In this study we want to demonstrate the degree of consistency in the active oddball task across subjects, in order to answer the question of whether this task is able to reliably detect conscious target processing in unresponsive patients. We asked 18 subjects to count rare targets and to ignore frequent standards and rare distractors in an auditory active oddball task. Event-related-potentials (ERPs) and time-frequency data were analyzed with permutation-t-tests on a single subject level. We plotted the group-average ERPs and time-frequency data, and evaluated the numbers of subjects showing significant differences between targets and distractors in certain time-ranges. The distinction between targets/distractors and standards was found to be significant in the time-range of the P300 in all participants. In contrast, significant differences between targets and distractors in the time-range of the P3a/b were found in 8 subjects, only. By including effects in the N1 and in a late negative component there remained 2 subjects who did not show a distinction between targets and distractors in the ERP. While time-frequency data showed prominent effects for target/distractor vs. standard, significant differences between targets and distractors were found in 2 subjects, only. The results suggest that time-frequency- and ERP-analysis of the active oddball task may not be sensitive enough to detect voluntary brain activation in unresponsive patients. In addition, we found that time-frequency analysis was even less informative than ERPs about the subject’s task performance. Despite suggesting the use of more sensitive paradigms and/or analysis techniques, the present results give further evidence that electroencephalographic research should rely more strongly on single-subject analysis because interpretations of group-effects may be misleading.

Reliability of EEG Measures of Interaction: A Paradigm Shift Is Needed to Fight the Reproducibility Crisis

Measures of interaction (connectivity) of the EEG are at the forefront of current neuroscientific research. Unfortunately, test-retest reliability can be very low, depending on the measure and its estimation, the EEG-frequency of interest, the length of the signal, and the population under investigation. In addition, artifacts can hamper the continuity of the EEG signal, and in some clinical situations it is impractical to exclude artifacts. We aimed to examine factors that moderate test-retest reliability of measures of interaction. The study involved 40 patients with a range of neurological diseases and memory impairments (age median: 60; range 21–76; 40% female; 22 mild cognitive impairment, 5 subjective cognitive complaints, 13 temporal lobe epilepsy), and 20 healthy controls (age median: 61.5; range 23–74; 70% female). We calculated 14 measures of interaction based on the multivariate autoregressive model from two EEG-recordings separated by 2 weeks. We characterized test-retest reliability by correlating the measures between the two EEG-recordings for variations of data length, data discontinuity, artifact exclusion, model order, and frequency over all combinations of channels and all frequencies, individually for each subject, yielding a correlation coefficient for each participant. Excluding artifacts had strong effects on reliability of some measures, such as classical, real valued coherence (~0.1 before, ~0.9 after artifact exclusion). Full frequency directed transfer function was highly reliable and robust against artifacts. Variation of data length decreased reliability in relation to poor adjustment of model order and signal length. Variation of discontinuity had no effect, but reliabilities were different between model orders, frequency ranges, and patient groups depending on the measure. Pathology did not interact with variation of signal length or discontinuity. Our results emphasize the importance of documenting reliability, which may vary considerably between measures of interaction. We recommend careful selection of measures of interaction in accordance with the properties of the data. When only short data segments are available and when the signal length varies strongly across subjects after exclusion of artifacts, reliability becomes an issue. Finally, measures which show high reliability irrespective of the presence of artifacts could be extremely useful in clinical situations when exclusion of artifacts is impractical.

Variability Issues in Automated Hippocampal Segmentation: A Study on Out-of-the-Box Software and Multi-rater Ground Truth

In automated hippocampus segmentation, issues related to ground truth rater variability, subject variability and variability of software segmentation accuracy are investigated in the context of 3 publicly available, out-of-the-box software packages. Ground truth variability among three manual raters is controlled using a majority voting based label fusion scheme and observed subject variability underpins the importance of availability of large scale ground truth.

Pathology-Related Automated Hippocampus Segmentation Accuracy

Hippocampal segmentation accuracy of out-of-the-box software tools (FreeSurfer, AHEAD, BrainParser) is analysed wrt. potential variability in populations with different pathologies. Findings confirm variabilities wrt. different pathologies but also human rater ground truth and single pathologies exhibit significant variability as well.

LP29: Considering transcallosal coherence as a marker of consciousness

Results: The statistics did not reveal notable differences between MCSand UWS-patients. The comparison of DOC-patients and healthy subjects revealed statistical trends (p<0.05, uncorrected) in about 53.66% of the computed tests. We found an interesting pattern of common differences between DOC-patients and healthy subjects in a frequency of 6Hz. Conclusions: Transcallosal coherence of MCS- and UWS-patients seems not to differ systematically. Due to the differences of DOC-patients and healthy subjects, one might conclude that the impairment of DOC-patients is based on a common mechanism to a certain extent. The earlier mentioned interesting pattern of common differences might reflect a damage to frontal midline theta coherence which in turn might reflect an impairment of the working memory [3]. References: