Identification of hemifield single trial PVEP on the basis of generalized dynamic neural network classifiers

Application of Generalized Dynamic Neural Networks to Biomedical Data

This paper reviews the application of continuous recurrent neural networks with time-varying weights to pattern recognition tasks in medicine. A general learning algorithm based on Pontryagin's maximum principle is recapitulated, and possibilities of improving the generalization capabilities of these networks are given. The effectiveness of the methods is demonstrated by three different real-world examples taken from the fields of anesthesiology, orthopedics, and radiology.

Classification of equinus in ambulatory children with cerebral palsy-discrimination between dynamic tightness and fixed contracture

In this study a generalised dynamic neural network (GDNN) was designed to process gait analysis parameters to evaluate equinus deformity in ambulatory children with cerebral palsy. The aim was to differentiate dynamic calf muscle tightness from fixed muscle contracture. Patients underwent clinical examination and had instrumented gait analysis before evaluating their equinus under anaesthesia and muscle relaxation at the time of surgery to improve gait. The performance of the clinical examination, the subjective interpretation of gait analysis results, and the application of the neural network to assess ankle function were compared to the examination under anaesthesia. Evaluation of equinus by a Neural Network showed high sensitivity and specificity values with a likelihood ratio of +14.63. The results indicate that dynamic calf muscle tightness can be differentiated from fixed calf muscle contracture with considerable precision that might facilitate clinical decision-making.

Prediction of movement following noxious stimulation during 1 minimum alveolar anesthetic concentration isoflurane/nitrous oxide anesthesia by means of middle latency auditory evoked responses

This paper investigates the applicability of generalized dynamic neural networks for the design of a two-valued anesthetic depth indicator during isoflurane/nitrous oxide anesthesia. The indicator construction is based on the processing of middle latency auditory evoked responses (MLAER) in combination with the observation of the patient's movement reaction to skin incision. The framework of generalized dynamic neural networks does not require any data preprocessing, visual data inspection or subjective feature extraction. The study is based on a data set of 106 patients scheduled for elective surgery under isoflurane/nitrous oxide anesthesia. The processing of the measured MLAER is performed by a recurrent neural network that transforms the MLAER signals into signals having a very uncomplex structure. The evaluation of these signals is self-evident, and yields to a simple threshold classifier. Using only evoked potentials before the pain stimulus, the patient's reaction could be predicted with a probability of 81.5%. The MLAER is closely associated to the patient's reaction to skin incision following noxious stimulation during 1 minimum alveolar anesthetic concentration isoflurane/nitrous oxide anesthesia. In combination with other parameters, MLAER could contribute to an objective and trustworthy movement prediction to noxious stimulation.

Electroencephalogram processing using neural networks

The electroencephalogram (EEG), a highly complex signal, is one of the most common sources of information used to study brain function and neurological disorders. More than 100 current neural network applications dedicated to EEG processing are presented. Works are categorized according to their objective (sleep analysis, monitoring anesthesia depth, brain-computer interface, EEG artifact detection, EEG source-based localization, etc.). Each application involves a specific approach (long-term analysis or short-term EEG segment analysis, real-time or time delayed processing, single or multiple EEG-channel analysis, etc.), for which neural networks were generally successful. The promising performances observed are demonstrative of the efficiency and efficacy of systems developed. This review can aid researchers, clinicians and implementors to understand up-to-date interest in neural network tools for EEG processing. The extended bibliography provides a database to assist in possible new concepts and idea development.

Identification of the stimulated hemiretina in primary school children and adults based on left and right hemifield pattern reversal visual evoked potentials--a comparative study

OBJECTIVES

The analysis of left and right hemifield pattern-reversal visual evoked potentials (PVEPs) in children and the identification of the stimulated hemiretina testing different identification procedures previously applied to adults.

METHODS

Lateral hemifield PVEPs were recorded in 40 children (6-11 years) and 27 adults (25-40 years) from, at least, 19 standard electrodes. Two procedures were tested for the determination of the stimulated hemifield: firstly, the evaluation of the values of instantaneous frequency at the occipital electrodes at P100 latency (determined by the global field power), and secondly, the application of a generalised dynamic neural network (GDNN) using the PVEP time course at selected electrode positions as the external input.

RESULTS

P100 latency as well as P100 amplitude over the contralateral occiput in children were significantly greater than in adults. Contrary to the behaviour in adults, instantaneous frequency is not a robust identifier of left and right hemiretina stimulation in children. The best identification performances were achieved when using group trained GDNNs with the bipolar difference signals of electrodes P3/P4 or T5/T6 as the external input.

CONCLUSIONS

The PVEPs at electrodes P3/P4 and T5/T6 contain essential information for the determination of the stimulated hemifield. This should be further considered during the development of on-line procedures for automatic PVEP detection in future studies.