Artificial intelligence methods in quantitative electroencephalogram analysis

Context related artefact detection in prolonged EEG recordings

The need for reliable detection of artefacts in raw and processed EEG is widely acknowledged. Although different EEG analysis systems have been described, only few general applicable artefact recognition techniques have emerged. This paper tackles the problem of artefact detection in seven 24 h EEG recordings in the intensive care unit. ICU recordings have received less attention than, e.g. epilepsy monitoring, although recordings in this environment present an interesting application area. The EEG data used here was recorded during the difficult circumstances of an explorative ICU study. The data set includes a diverse set of EEG patterns, as well as EEG artefacts. The study investigates objective artefact detection methods based on statistical differences between signal parameters, using time-varying autoregressive modelling (AR) and Slope detection. In addition to matching the performance of artefact detection against two human observers, the study focuses on the optimal settings for context incorporation by testing the algorithms for different time windows and epoch lengths. Results indicate that a relatively short period (20-40 s) provides sufficient context information for the methods used. The combined AR and Slope detection parameters yielded good performance, detecting approximately 90% of the artefacts as indicated by the consensus score of the human observers.

Automated staging of sleep in cats using neural networks

Manual staging of sleep based on visual EEG criteria is a laborious and time-consuming task. In an effort to automate sleep staging, we have developed a neural network that 'learns' to stage sleep on the basis of wave band count data alone, in the cat. Wave band count data are collected on a microcomputer, using period-amplitude analysis. Delta waves, spindle bursts, ponto-geniculo-occipital (PGO) waves, electro-oculogram (EOG), basal electromyogram (EMG) amplitude, and movement artifact amplitude are collected, and used to 'train' the network to score sleep. These wave count data serve as the input patterns to the net, and the corresponding manually scored sleep stages serve as a 'teacher.' We demonstrate that, when used to score the states of wake, slow wave sleep (SWS), desynchronized sleep (D), and the transition period from SWS to D (SP), these neural networks agree with manual scoring an average of 93.3% for all epochs scored. Neural network programs can learn both rules and exceptions, and since the nets teach themselves these rules automatically, a minimum of human effort is required. Because programming requirements are small for neural nets, this approach is readily adaptable to microcomputer-based systems and is widely applicable to both animal and human EEG analyses. The utility of this approach for the detection and classification of a variety of clinical neurophysiological disorders is discussed.

Quantitative analysis of electroencephalograms: is there chaos in the future?

The history of quantitative, computerized electroencephalogram (EEG) analysis is reviewed. It is shown that, until very recently, the basic approach to EEG analysis involved the assumption that the EEG is stochastic. Consequently, statistical pattern recognition techniques, segmentation procedures, syntactic methods, knowledge-based approaches, and even artificial neural network methods have been developed with different levels of success. A fundamentally different approach to computerized EEG analysis, however, is making its way into the laboratories. The basic idea, inspired by recent advances in the area of non-linear dynamics, and especially the theory of chaos, is to view an EEG as the output of a deterministic system of relatively simple complexity, but containing non-linearities. This suggests that studying the geometrical dynamics of EEGs, and the development of neurophysiologically realistic models of EEG generation may produce more successful automated EEG analysis techniques than the classical, stochastic methods. Evidence supporting the non-linear dynamics paradigm is reviewed, and possible research paths are indicated.

Knowledge-based approach to sleep EEG analysis--a feasibility study

A knowledge-based approach to automated sleep EEG (electroencephalogram) analysis is described. In this system, an object-oriented approach is followed in which specific waveforms and sleep stages ("objects") are represented in terms of frames. The latter capture the morphological and spatio-temporal information for each object. An object detection module ("frame matcher"), operating on the frames, is employed to identify what features need to be extracted from the EEG and to trigger the appropriate "specialist"--specialized signal processing modules--to obtain values for these features. This leads to an opportunistic approach to EEG interpretation with quantitative information being extracted from the signal only when needed by the reasoning processes. The system has been tested on the detection of K complexes and sleep spindles. Its performance indicates that the approach followed is feasible and can become a powerful tool for automated EEG interpretation.

Elements of EEG signal processing

It is difficult to find well controlled clinical studies comparing the utility of the numerous EEG measures which have been described. A brief summary of the literature reveals a number of promising case reports, a few large series of patients, and fewer well-designed, well-controlled studies. The very abundance of algorithms makes even technical comparisons of the ability of each to transduce and reflect clinical changes useful (27, 29). Most studies are positive; but thus far (as with so many procedures in medicine), there is no definitive study demonstrating an unequivocal benefit to patients who are monitored with automated, online EEG analysis. That such a study may never be done should not detract from the possible benefits these techniques may bring to clinical practice, most particularly preparing the way for effective brain monitoring in situations in which it was impractical before.

Computerized EEG pattern classification by adaptive segmentation and probability density function classification. Clinical evaluation

A series of 63 clinical EEGs showing a variety of normal and abnormal patterns was analysed by computer with particular reference to the different types of pattern within the same EEG. Boundaries between different patterns were established by means of adaptive segmentation, so that the duration of the resulting segments was determined by the particular EEG itself (thus the term 'adaptive'). Four channels from each EEG were analysed, paired (left and right) channels were simultaneously segmented and analysed interactively. Similar segments were then clustered without supervision by estimating a probability density function in a 2-dimensional 'feature space' having dimensions of mean frequency and mean power. Individual clusters emerged as well-defined peaks of the surface, individual segments or small groups of duration insufficient to constitute a separate cluster, being identified as 'singular events' (e.g., rare sharp waves, artifacts). The autocorrelation function was used to characterize the EEG both for the segmentation and for the subsequent clustering of the resulting segments. In confirmation of our previous work, adaptive segmentation based on the autocorrelation function of the EEG was found to be quite satisfactory. Unsupervised clustering by estimation of the probability density function in feature space was found to give the correct number of clusters (usually less than 5) in a majority of the records (65%), but in the remaining minority of cases (35%), either overclustering or underclustering occurred. Further, the 'singular events' were occasionally partly included in a formal cluster. Comparison of these results of EEG clustering by unsupervised probability density function estimation with earlier results obtained by supervised hierarchical clustering suggests that there may be subtle cues used by the electroencephalographer in the classification of EEG patterns which have not been adequately approximated by the computer algorithms thus far used in this work. Hence at least some minimal degree of supervision in the clustering process may be necessary, at least for the present. On the other hand, the method recommends itself for the representation of illustrative EEG summaries which, in conjunction with a short written report, would provide the clinical neurologist with a sufficient picture of the real EEG without, in most cases, the need to inspect the original record.