Robert Keith

A multicentre comparative study of 17 experts and an intelligent computer system for managing labour using the cardiotocogram

Objectives To investigate 1. whether an intelligent computer system could obtain a performance in labour management comparable with experts when using cardiotocograms (CTGs), patient information, and fetal blood sampling and 2. whether experts could be consistent and agree in their management of labour.

4 Development, evaluation and validation of an intelligent system for the management of labour

Over the past 4 years our group has developed a prototype intelligent system which applies captured expert knowledge to support clinical decision-making during labour. This chapter presents a review of the system and the progress made to date. The system classifies the same features from the CTG as experienced clinicians using numerical algorithms and a small neural network. This hybrid approach has been shown to obtain a comparable performance with experts. The CTG information, together with the patient information and labour events, are collectively passed to an expert system for processing. The expert system interprets this combined data using a database of over 400 rules which are used to recommend action. Importantly, as the knowledge is rule-based, it allows the system to explain the reasoning which led it to recommend a certain action. In this way, the clinician is not expected to blindly follow the systems recommendations but can reach an informed judgement in the same way they might by discussing the case with an experienced informed colleague. After two internal evaluations had found the system obtained a performance comparable with local experts, an extensive external validation was undertaken. This study involved 17 experts from 16 leading centres within the UK. Each expert and the system reviewed 50 cases twice, at least one month apart which contained those CTGs considered most difficult to interpret selected from a database of 2400 high-risk labours. This study found that the majority of experts agreed well and were consistent in their management of the cases. The system obtained a performance that was indistinguishable from the experts, except it was more consistent, even when used by an engineer with little knowledge of labour management. This study demonstrates the potential for intelligent systems to transform the cardiotocograph from a difficult-to-use, ineffective recorder of fetal heart rate, to an interactive and effective decision support tool capable of raising the skills of staff.

Suitability of fetal scalp electrodes for monitoring the fetal electrocardiogram during labour

As the limitations of heart-rate based intrapartum monitoring have become apparent, there is renewed interest in analysis of the fetal electrocardiographic waveform as obtained from a fetal scalp electrode. A high quality ECG signal is necessary for waveform analysis. This study examined the suitability of five commonly available scalp electrodes for collecting this signal by examining their physical and electrical characteristics, together with a randomised clinical trial in which the ECG trace quality was assessed in 50 patients. The frequency response of Copeland electrodes was such that they attenuate the ECG signal more than the baseline noise. Difficulties were experienced in obtaining optimum attachment and the long, semi-rigid design increased movement artefact resulting in significantly poorer quality ECG signals. Whilst the Hewlett-Packard double spiral electrode had a near ideal frequency response, certain design features made it difficult to apply and remain secure so the clinical signals were of intermediate quality. The Corometrics and Cetro single spirals had the most stable attachment to the scalp and a near ideal frequency response, so produced significantly better signal quality in the clinical trial. Currently, single spiral electrodes are the most suitable for electrocardiographic data collection.

Suitability of artificial neural networks for feature extraction from cardiotocogram during labour

Fetal condition during labour is inferred from a continuous display of fetal heartrate and uterine contractions called the cardiotocogram (CTG). The CTG requires a considerable expertise for correct interpretation, which is not always available. We are developing an intelligent system to support clinical decision-making during labour. The system’s performance depends on its ability to classify features from the CTG similarly to experts. Artificial neural networks 9NNs) can be taught by experts for such tasks, and so may be particularly suitable. We found NNs suitable for feature extraction when the problem was reduced to small well defined tasks, and numerical algorithms were used to pre-process the raw data before application to the NNs. A NN with optimised dimensions was used in this way to classify the magnitude of decelerations, a feature clinicians find particularly difficult. The NN was compared with the algorithm used in a commercial antenatal monitor* and six reviewers which included two CTG experts. The experts were consistent (89·7% and 97·0%) and agreed well with each other (81·0%), whereas the non-experts were less consistent and agreed less well. The NN agreed well with the experts (75·0% and 81·9%) but the algorithm agreed poorly (56.5% and 68·9%). It was found that the algorithm’s performance could be improved (71·1% and 76·7%) when modified to use additional information. Our earlier attempts to fully classify the raw CTG using a single NN were unsuccessful because of the large number of data patterns. A simplified approach to classify the magnitude and timing of decelerations was also unsuitable when contraction data was of poor quality or absent. We have adopted a hybrid approach for our intelligent system for reliability and improved performance. CTG features are extracted and classified by combined numerical algorithms and small NNs. These features, together with patient information, are then processed by an expert system which allows interaction with clinicians.

Umbilical cord blood gas analysis at the time of delivery

AIMS it is now recommended that cord blood acid-base measurement is performed routinely at time of delivery in the UK as a measure of fetal response to labour. However, there remains some uncertainty about the value of this procedure. In this paper our experience of cord blood analysis is described and the literature is reviewed to: (1) provide an overview of the physiological basis of cord blood acid-base assessment; (2) describe the appropriate methodology and identify issues which have contributed to confusion and undermined the value of cord blood sampling; and (3) address the practical issues of cord blood sampling. CONCLUSIONS cord blood acid-base measurement has a sound physiological basis. It provides objective information which is a useful adjunct to subjective methods of newborn assessment, enables babies at risk of neonatal morbidity to be identified, can be helpful in litigation cases and is a prerequisite for clinical audit. However, to be of benefit the information must be correct and correctly interpreted.

Multicentre Validation of an Intelligent System for Managing Labour

Fetal condition during labour is inferredfrom a continuous visual recording of the fetal heart rate and uterine contractions, called the cardiotocogram (CTG). DifJiculties in CTG interpretation is a major problem that can lead to both unnecessary Caesarean sections and damage to the infant. Our group has developed an intelligent system which applies expert knowledge to support clinical decision making during laboul: The system was previously found to obtain a performance comparable to local experts in two internal evaluations. This study presents a validation of the system which compared its management with I7 experts from 16 leading centres within the U.K. in 50 cases selectedfrom a database of 2400 high risk labours. This study found that the majority of experts agreed well and were largely consistent in their management of the cases. The system obtained a performance that was indistinguishable from the experts. Copyright 0 1996 Elsevier Science Ltd

The prediction of fetal acidosis at birth by computerised analysis of intrapartum cardiotocography

Sir, We would like to thank Dr Pamela J Schwingl for giving us the opportunity to clarify a phrase in our paper which may have been misinterpreted by readers other than Dr Schwingl. When we indicated that the odds ratio (= relative risk for rare diseases) of ischaemic stroke among hypertensive women was independent of combined oral contraceptive (COC) status, this indicated that hypertension implied the same odds ratio among women on OC as among women not using COC. In a multiplicative model this independence means that the exposure of two risk factors implies a risk of stroke corresponding a multiplication of the two risk factors. As low dose COCs (30-40 Hg oestrogen) in themselves imply a relative risk of stroke of 1.8 and hypertension a relative risk of about 3.1, the combined relative risk is close to 5.6. All the included risk factors were analyzed for interaction with oral contraceptive use. No evidence of interaction was found. In general, women with a known increased risk of stroke should be advised against increasing their risk any further by adding other risk factors. There is no contradiction in showing an independence between two risk factors and at the same time advising women with an increased risk of stroke from increasing that risk further by, for example, COC use. We are well aware that modern epidemiology has demonstrated that two biological exposures acting at the same time may result in a total risk that corresponds to an addition of the two relative risks, rather than a multiplication of the two risks factors. In the example of oral contraceptives and hypertension an additive combined effect would result in a combined relative risk of 4.9, implying an interaction between the two exposures in a multiplicative model. Several different test statistics were used, all resulting in p-values between 0.3 and 08. As an analysis of the differences between the observed and fitted values (residuals) showed a good fit, we see no problem in accepting the multiplicative model as a realistic description of the combined effect of the included risk factors. In an additive model this would of course imply an interaction which one might interpret as synergy.