Computer analysis of antepartum fetal heart rate: 2. Detection of accelerations and decelerations

Fetal states identification in cardiotocographic tracings through discrete emissions multivariate hidden Markov models

BACKGROUND AND OBJECTIVES

Computerized Cardiotocography (cCTG) allows to analyze the Fetal Heart Rate (FHR) objectively and thoroughly, providing valuable insights on fetal condition. A challenging but crucial task in this context is the automatic identification of fetal activity and quiet periods within the tracings. Different neural mechanisms are involved in the regulation of the fetal heart, depending on the behavioral states. Thereby, their correct identification has the potential to increase the interpretability and diagnostic capabilities of FHR quantitative analysis. Moreover, the most common pathologies in pregnancy have been associated with variations in the alternation between quiet and activity states.

METHODS

We address the problem of fetal states clustering by means of an unsupervised approach, resorting to the use of a multivariate Hidden Markov Models (HMM) with discrete emissions. A fixed length sliding window is shifted on the CTG traces and a small set of features is extracted at each slide. After an encoding procedure, these features become the emissions of a multivariate HMM in which quiet and activity are the hidden states. After an unsupervised training procedure, the model is used to automatically segment signals.

RESULTS

The achieved results indicate that our developed model exhibits a high degree of reliability in identifying quiet and activity states within FHR signals. A set of 35 CTG signals belonging to different pregnancies were independently annotated by an expert gynecologist and segmented using the proposed HMM. To avoid any bias, the physician was blinded to the results provided by the algorithm. The overall agreement between the HMM's predictions and the clinician's interpretations was 90%.

CONCLUSIONS

The proposed method reliably identified fetal behavioral states, the alternance of which is an important factor in the fetal development. One key strength of our approach lies in the ease of interpreting the obtained results. By utilizing a small set of parameters that are already used in cCTG and possess clear intrinsic meanings, our method provides a high level of explainability. Another significant advantage of our approach is its fully unsupervised learning process. The states identified by our model using the Baum-Welch algorithm are associated with the "Active" and "Quiet" states only after the clustering process, removing the reliance on expert annotations. By autonomously identifying the clusters based solely on the intrinsic characteristics of the signal, our method achieves a more objective evaluation that overcomes the limitations of subjective interpretations. Indeed, we believe it could be integrated in cCTG systems to obtain a more complete signal analysis.

Fetal heart rate baseline computation with a weighted median filter

BACKGROUND

Automated fetal heart rate (FHR) analysis removes inter- and intra-expert variability, and is a promising solution for reducing the occurrence of fetal acidosis and the implementation of unnecessary medical procedures. The first steps in automated FHR analysis are determination of the baseline, and detection of accelerations and decelerations (A/D). We describe a new method in which a weighted median filter baseline (WMFB) is computed and A/Ds are then detected.

METHOD

The filter weightings are based on the prior probability that the sampled FHR is in the baseline state or in an A/D state. This probability is computed by estimating the signal's stability at low frequencies and by progressively trimming the signal. Using a competition dataset of 90 previously annotated FHR recordings, we evaluated the WMFB method and 11 recently published literature methods against the ground truth of an expert consensus. The level of agreement between the WMFB method and the expert consensus was estimated by calculating several indices (primarily the morphological analysis discordance index, MADI). The agreement indices were then compared with the values for eleven other methods. We also compared the level of method-expert agreement with the level of interrater agreement.

RESULTS

For the WMFB method, the MADI indicated a disagreement of 4.02% vs. the consensus; this value is significantly lower (p<10^-13) than that calculated for the best of the 11 literature methods (7.27%, for Lu and Wei's empirical mode decomposition method). The level of inter-expert agreement (according to the MADI) and the level of WMFB-expert agreement did not differ significantly (p=0.22).

CONCLUSION

The WMFB method reproduced the expert consensus analysis better than 11 other methods. No differences in performance between the WMFB method and individual experts were observed. The method Matlab source code is available under General Public Licence at http://utsb.univ-catholille.fr/fhr-wmfb.

Computer-Aided Diagnosis System of Fetal Hypoxia Incorporating Recurrence Plot With Convolutional Neural Network

Background: Electronic fetal monitoring (EFM) is widely applied as a routine diagnostic tool by clinicians using fetal heart rate (FHR) signals to prevent fetal hypoxia. However, visual interpretation of the FHR usually leads to significant inter-observer and intra-observer variability, and false positives become the main cause of unnecessary cesarean sections. Goal: The main aim of this study was to ensure a novel, consistent, robust, and effective model for fetal hypoxia detection. Methods: In this work, we proposed a novel computer-aided diagnosis (CAD) system integrated with an advanced deep learning (DL) algorithm. For a 1-dimensional preprocessed FHR signal, the 2-dimensional image was transformed using recurrence plot (RP), which is considered to greatly capture the non-linear characteristics. The ultimate image dataset was enriched by changing several parameters of the RP and was then used to feed the convolutional neural network (CNN). Compared to conventional machine learning (ML) methods, a CNN can self-learn useful features from the input data and does not perform complex manual feature engineering (i.e., feature extraction and selection). Results: Finally, according to the optimization experiment, the CNN model obtained the average performance using optimal configuration across 10-fold: accuracy = 98.69%, sensitivity = 99.29%, specificity = 98.10%, and area under the curve = 98.70%. Conclusion: To the best of our knowledge, this approached achieved better classification performance in predicting fetal hypoxia using FHR signals compared to the other state-of-the-art works. Significance: In summary, the satisfied result proved the effectiveness of our proposed CAD system for assisting obstetricians making objective and accurate medical decisions based on RP and powerful CNN algorithm.

Prognostic model based on image-based time-frequency features and genetic algorithm for fetal hypoxia assessment

Cardiotocography (CTG) is applied routinely for fetal monitoring during the perinatal period to decrease the rates of neonatal mortality and morbidity as well as unnecessary interventions. The analysis of CTG traces has become an indispensable part of present clinical practices; however, it also has serious drawbacks, such as poor specificity and variability in its interpretation. The automated CTG analysis is seen as the most promising way to overcome these disadvantages. In this study, a novel prognostic model is proposed for predicting fetal hypoxia from CTG traces based on an innovative approach called image-based time-frequency (IBTF) analysis comprised of a combination of short time Fourier transform (STFT) and gray level co-occurrence matrix (GLCM). More specifically, from a graphical representation of the fetal heart rate (FHR) signal, the spectrogram is obtained by using STFT. The spectrogram images are converted into 8-bit grayscale images, and IBTF features such as contrast, correlation, energy, and homogeneity are utilized for identifying FHR signals. At the final stage of the analysis, different subsets of the feature space are applied as the input to the least square support vector machine (LS-SVM) classifier to determine the most informative subset. For this particular purpose, the genetic algorithm is employed. The prognostic model was performed on the open-access intrapartum CTU-UHB CTG database. The sensitivity and specificity obtained using only conventional features were 57.33% and 67.24%, respectively, whereas the most effective results were achieved using a combination of conventional and IBTF features, with a sensitivity of 63.45% and a specificity of 65.88%. Conclusively, this study provides a new promising approach for feature extraction of FHR signals. In addition, the experimental outcomes showed that IBTF features provided an increase in the classification accuracy.

Machine learning ensemble modelling to classify caesarean section and vaginal delivery types using Cardiotocography traces

Human visual inspection of Cardiotocography traces is used to monitor the foetus during labour and avoid neonatal mortality and morbidity. The problem, however, is that visual interpretation of Cardiotocography traces is subject to high inter and intra observer variability. Incorrect decisions, caused by miss-interpretation, can lead to adverse perinatal outcomes and in severe cases death. This study presents a review of human Cardiotocography trace interpretation and argues that machine learning, used as a decision support system by obstetricians and midwives, may provide an objective measure alongside normal practices. This will help to increase predictive capacity and reduce negative outcomes. A robust methodology is presented for feature set engineering using an open database comprising 552 intrapartum recordings. State-of-the-art in signal processing techniques is applied to raw Cardiotocography foetal heart rate traces to extract 13 features. Those with low discriminative capacity are removed using Recursive Feature Elimination. The dataset is imbalanced with significant differences between the prior probabilities of both normal deliveries and those delivered by caesarean section. This issue is addressed by oversampling the training instances using a synthetic minority oversampling technique to provide a balanced class distribution. Several simple, yet powerful, machine-learning algorithms are trained, using the feature set, and their performance is evaluated with real test data. The results are encouraging using an ensemble classifier comprising Fishers Linear Discriminant Analysis, Random Forest and Support Vector Machine classifiers, with 87% (95% Confidence Interval: 86%, 88%) for Sensitivity, 90% (95% CI: 89%, 91%) for Specificity, and 96% (95% CI: 96%, 97%) for the Area Under the Curve, with a 9% (95% CI: 9%, 10%) Mean Square Error.

Classification of caesarean section and normal vaginal deliveries using foetal heart rate signals and advanced machine learning algorithms

BACKGROUND

Visual inspection of cardiotocography traces by obstetricians and midwives is the gold standard for monitoring the wellbeing of the foetus during antenatal care. However, inter- and intra-observer variability is high with only a 30% positive predictive value for the classification of pathological outcomes. This has a significant negative impact on the perinatal foetus and often results in cardio-pulmonary arrest, brain and vital organ damage, cerebral palsy, hearing, visual and cognitive defects and in severe cases, death. This paper shows that using machine learning and foetal heart rate signals provides direct information about the foetal state and helps to filter the subjective opinions of medical practitioners when used as a decision support tool. The primary aim is to provide a proof-of-concept that demonstrates how machine learning can be used to objectively determine when medical intervention, such as caesarean section, is required and help avoid preventable perinatal deaths.

METHODS

This is evidenced using an open dataset that comprises 506 controls (normal virginal deliveries) and 46 cases (caesarean due to pH ≤ 7.20-acidosis, n = 18; pH > 7.20 and pH < 7.25-foetal deterioration, n = 4; or clinical decision without evidence of pathological outcome measures, n = 24). Several machine-learning algorithms are trained, and validated, using binary classifier performance measures.

RESULTS

The findings show that deep learning classification achieves sensitivity = 94%, specificity = 91%, Area under the curve = 99%, F-score = 100%, and mean square error = 1%.

CONCLUSIONS

The results demonstrate that machine learning significantly improves the efficiency for the detection of caesarean section and normal vaginal deliveries using foetal heart rate signals compared with obstetrician and midwife predictions and systems reported in previous studies.

Software for computerised analysis of cardiotocographic traces

Despite the widespread use of cardiotocography in foetal monitoring, the evaluation of foetal status suffers from a considerable inter and intra-observer variability. In order to overcome the main limitations of visual cardiotocographic assessment, computerised methods to analyse cardiotocographic recordings have been recently developed. In this study, a new software for automated analysis of foetal heart rate is presented. It allows an automatic procedure for measuring the most relevant parameters derivable from cardiotocographic traces. Simulated and real cardiotocographic traces were analysed to test software reliability. In artificial traces, we simulated a set number of events (accelerations, decelerations and contractions) to be recognised. In the case of real signals, instead, results of the computerised analysis were compared with the visual assessment performed by 18 expert clinicians and three performance indexes were computed to gain information about performances of the proposed software. The software showed preliminary performance we judged satisfactory in that the results matched completely the requirements, as proved by tests on artificial signals in which all simulated events were detected from the software. Performance indexes computed in comparison with obstetricians' evaluations are, on the contrary, not so satisfactory; in fact they led to obtain the following values of the statistical parameters: sensitivity equal to 93%, positive predictive value equal to 82% and accuracy equal to 77%. Very probably this arises from the high variability of trace annotation carried out by clinicians.

Monitoring fetal heart rate during pregnancy: contributions from advanced signal processing and wearable technology

Monitoring procedures are the basis to evaluate the clinical state of patients and to assess changes in their conditions, thus providing necessary interventions in time. Both these two objectives can be achieved by integrating technological development with methodological tools, thus allowing accurate classification and extraction of useful diagnostic information. The paper is focused on monitoring procedures applied to fetal heart rate variability (FHRV) signals, collected during pregnancy, in order to assess fetal well-being. The use of linear time and frequency techniques as well as the computation of non linear indices can contribute to enhancing the diagnostic power and reliability of fetal monitoring. The paper shows how advanced signal processing approaches can contribute to developing new diagnostic and classification indices. Their usefulness is evaluated by comparing two selected populations: normal fetuses and intra uterine growth restricted (IUGR) fetuses. Results show that the computation of different indices on FHRV signals, either linear and nonlinear, gives helpful indications to describe pathophysiological mechanisms involved in the cardiovascular and neural system controlling the fetal heart. As a further contribution, the paper briefly describes how the introduction of wearable systems for fetal ECG recording could provide new technological solutions improving the quality and usability of prenatal monitoring.

Adaptive pattern recognition in the analysis of cardiotocographic records

The recognition of accelerative and decelerative patterns in the fetal heart rate (FHR) is one of the tasks carried out manually by obstetricians when they analyze cardiotocograms for information respecting the fetal state. An approach based on artificial neural networks formed by a multilayer perceptron (MLP) is developed. However, since the system utilizes the FHR signal as direct input, an anterior stage must be incorporated that applies a principal component analysis (PCA) so as to make the system independent of the signal baseline. Furthermore, the introduction of multiresolution into the PCA has resolved other problems that were detected in the application of the system. Presented in this paper are the results of validation of these systems designated the PCA-MLP and multiresolutlon principal component analysis (MR-PCA) systems against three clinical experts.

Simulation of foetal phonocardiographic recordings for testing of FHR extraction algorithms

A valuable alternative to traditional diagnostic tools, such as ultrasonographic cardiotocography, to monitor general foetal well-being by means of foetal heart rate analysis is foetal phonocardiography, a passive and low cost recording of foetal heart sounds. In this paper, it is presented a simulator software of foetal phonocardiographic signals relative to different foetal states and recording conditions (for example different kinds and levels of noise). Before developing the software, a data collection pilot study was conducted with the purpose of specifically identifying the characteristics of the waveforms of the foetal and maternal heart sounds, since the available literature is not rigorous in this area. The developed software, due to the possibility to simulate different physiological and pathological foetal conditions and recording situations simply modifying some system parameters, can be useful as a teaching tool for demonstration to medical students and others and also for testing and assessment of foetal heart rate extraction algorithms from foetal phonocardiographic (fPCG) recordings. On this purpose, the simulator software was used to test an algorithm developed by the authors for foetal heart rate extraction considering different foetal heart rate parameters and signal to noise ratio values. Our tests demonstrated that simulated fPCG signals are very close to real fPCG recordings.

FETAL HEART RATE VARIABILITY ANALYSIS FROM PHONOCARDIOGRAPHIC RECORDINGS

This paper presents an algorithm for classification of fetal health status using fetal heart rate variability (fHRV) analysis through phonocardiography. First, the fetal heart sound signals are acquired from the maternal abdominal surface using a specially developed Bluetooth-based wireless data recording system. Then, fetal heart rate (FHR) traces are derived from these signals. Ten numbers of linear and nonlinear features are extracted from each FHR trace. Finally, the multilayer perceptron (MLP) neural network is used to classify the health status of the fetus. Results show very promising performance toward the prediction of fetal wellbeing on the set of collected fetal heart sound signals. Finally, this work is likely to lead to an automatic screening device with additional potential of predicting fetal wellbeing.

FEATURE EXTRACTION AND CLASSIFICATION OF FETAL HEART RATE USING WAVELET ANALYSIS AND SUPPORT VECTOR MACHINES

Since the fetus is not available for direct observations, only indirect information can guide the obstetrician in charge. Electronic Fetal Monitoring (EFM) is widely used for assessing fetal well being. EFM involves detection of the Fetal Heart Rate (FHR) signal and the Uterine Activity (UA) signal. The most serious fetal incident is the hypoxic injury leading to cerebral palsy or even death, which is a condition that must be predicted and avoided. This research work proposes a new integrated method for feature extraction and classification of the FHR signal able to associate FHR with umbilical artery pH values at delivery. The proposed method introduces the use of the Discrete Wavelet Transform (DWT) to extract time-scale dependent features of the FHR signal and the use of Support Vector Machines (SVMs) for the categorization. The proposed methodology is tested on a data set of intrapartum recordings were the FHR categories are associated with umbilical artery pH values, This proposed approach achieved high overall classification performance proving its merits.

Antepartum fetal heart rate feature extraction and classification using empirical mode decomposition and support vector machine

Background

Cardiotocography (CTG) is the most widely used tool for fetal surveillance. The visual analysis of fetal heart rate (FHR) traces largely depends on the expertise and experience of the clinician involved. Several approaches have been proposed for the effective interpretation of FHR. In this paper, a new approach for FHR feature extraction based on empirical mode decomposition (EMD) is proposed, which was used along with support vector machine (SVM) for the classification of FHR recordings as 'normal' or 'at risk'.

Methods

The FHR were recorded from 15 subjects at a sampling rate of 4 Hz and a dataset consisting of 90 randomly selected records of 20 minutes duration was formed from these. All records were labelled as 'normal' or 'at risk' by two experienced obstetricians. A training set was formed by 60 records, the remaining 30 left as the testing set. The standard deviations of the EMD components are input as features to a support vector machine (SVM) to classify FHR samples.

Results

For the training set, a five-fold cross validation test resulted in an accuracy of 86% whereas the overall geometric mean of sensitivity and specificity was 94.8%. The Kappa value for the training set was .923. Application of the proposed method to the testing set (30 records) resulted in a geometric mean of 81.5%. The Kappa value for the testing set was .684.

Conclusions

Based on the overall performance of the system it can be stated that the proposed methodology is a promising new approach for the feature extraction and classification of FHR signals.

The application of empirical mode decomposition for the enhancement of cardiotocograph signals

Cardiotocograph (CTG) is widely used in everyday clinical practice for fetal surveillance, where it is used to record fetal heart rate (FHR) and uterine activity (UA). These two biosignals can be used for antepartum and intrapartum fetal monitoring and are, in fact, nonlinear and non-stationary. CTG recordings are often corrupted by artifacts such as missing beats in FHR, high-frequency noise in FHR and UA signals. In this paper, an empirical mode decomposition (EMD) method is applied on CTG signals. A recursive algorithm is first utilized to eliminate missing beats. High-frequency noise is reduced using EMD followed by the partial reconstruction (PAR) method, where the noise order is identified by a statistical method. The obtained signal enhancement from the proposed method is validated by comparing the resulting traces with the output obtained by applying classical signal processing methods such as Butterworth low-pass filtering, linear interpolation and a moving average filter on 12 CTG signals. Three obstetricians evaluated all 12 sets of traces and rated the proposed method, on average, 3.8 out of 5 on a scale of 1(lowest) to 5 (highest).

Long-term Phonocardiographic Fetal Home Monitoring for Telemedicine Systems

A novel compact method for fetal home monitoring optimized for long data acquisition time and low communication costs is presented. The method incorporates the preprocessing of disturbed acoustic signal received on the maternal abdomen. The basic idea of the preprocessing is that the detection of the systolic and diastolic sounds takes place on two separated frequency bands with autocorrelation on predicted time intervals. Measurements on 47 selected pregnant women have shown that the use of this method significantly reduces the amount of data to be transferred to the computer centre in the hospital, where only the very disturbed time periods have to be evaluated. Based on this method a new, phonocardiographic fetal telemedicine system can be built without time limitation of measurements.

Predicting the risk of metabolic acidosis for newborns based on fetal heart rate signal classification using support vector machines

Cardiotocography is the main method used for fetal assessment in every day clinical practice for the last 30 years. Many attempts have been made to increase the effectiveness of the evaluation of cardiotocographic recordings and minimize the variations of their interpretation utilizing technological advances. This research work proposes and focuses on an advanced method able to identify fetuses compromised and suspicious of developing metabolic acidosis. The core of the proposed method is the introduction of a support vector machine to "foresee" undesirable and risky situations for the fetus, based on features extracted from the fetal heart rate signal at the time and frequency domains along with some morphological features. This method has been tested successfully on a data set of intrapartum recordings, achieving better and balanced overall performance compared to other classification methods, constituting, therefore, a promising new automatic methodology for the prediction of metabolic acidosis.

Foetal heart rate power spectrum response to uterine contraction

Cardiotocography is the most diffused prenatal diagnostic technique in clinical routine. The simultaneous recording of foetal heart rate (FHR) and uterine contractions (UC) provides useful information about foetal well-being during pregnancy and labour. However, foetal electronic monitoring interpretation still lacks reproducibility and objectivity. New methods of interpretation and new parameters can further support physicians' decisions. Besides common time-domain analysis, study of the variability of FHR can potentially reveal autonomic nervous system activity of the foetus. In particular, it is clinically relevant to investigate foetal reactions to UC to diagnose foetal distress early. Uterine contraction being a strong stimulus for the foetus and its autonomic nervous system, it is worth exploring the FHR variability response. This study aims to analyse modifications of the power spectrum of FHR variability corresponding to UC. Cardiotocographic signal tracts corresponding to 127 UC relative to 30 healthy foetuses were analysed. Results mainly show a general, statistically significant (t test, p<0.01) power increase of the FHR variability in the LF 0.03-0.2 Hz and HF 0.2-1 in correspondence of the contraction with respect to a reference tract set before contraction onset. Time evolution of the power within these bands was computed by means of time-varying spectral estimation to concisely show the FHR response along a uterine contraction. A synchronised grand average of these responses was also computed to verify repeatability, using the contraction apex as time reference. Such modifications of the foetal HRV that follow a contraction can be a sign of ANS reaction and, therefore, additional, objective information about foetal reactivity during labour.

Integrated monitoring of fetal growth restriction by computerized cardiotocography and Doppler flow velocimetry

OBJECTIVE

To investigate the correlations between Doppler flow velocimetry and computerized cardiotocography (cCTG) in fetal growth restriction.

STUDY DESIGN

Fifty growth-restricted foetuses with abdominal circumference below the 10th percentile and no major abnormalities were studied. A total of 186 cCTG tracings (at least two per patient) analysed using the HP2CTG system were compared with the corresponding umbilical artery pulsatility index (PI), the PI ratio of umbilical artery to middle cerebral artery, and the ductus venosus systolic/atrial ratio.

RESULTS

Worsening in umbilical artery Doppler velocimetry parameters was associated with a significant reduction of short- and long-term variability indices and accelerations. When end-diastolic umbilical artery flow was preserved, a reversed ratio between umbilical artery and middle cerebral artery PIs was not correlated with a worsening of cCTG parameters; in the presence of umbilical artery absent or reversed flow, ductus venosus Doppler velocimetry abnormalities were correlated with a significant reduction of variability. When end-diastolic umbilical artery flow was preserved, there was a progressive increase in variability indices and accelerations with advancing gestational age. In the more compromised fetuses this "maturation" process of cCTG patterns was not found.

CONCLUSION

There is a strict correlation between Doppler velocimetry abnormalities and cCTG parameter deterioration, in particular between ductus venosus and variability.

Antepartum cardiotocography: a study of fetal reactivity in frequency domain

Cardiotocography (CTG) is the most widely used diagnostic technique in clinical practice to monitor fetal health. Cardiotocographic recording also permits to assess maturation of the fetal autonomous nervous system (ANS): fetal heart rate (FHR) modifications may reveal ANS' reactions to stimuli. To assess fetal reactivity, physicians evaluate specific clinical CTG parameters, generally, by means of visual inspection, thus depending on observer's expertise, with lack of reproducibility. Still nowadays, there is a very high intra- and inter-observer variation in the assessment of FHR patterns. More objective methods for CTG interpretation are of crucial importance. For adults, frequency analysis of heart rate variability (HRV) is a non-invasive and powerful method to investigate ANS activity. This frequency analysis can also be a valid support for a better knowledge of fetal ANS functional state and reactions. Indeed, fetal HRV is a good indicator of fetal well-being in non-stress conditions. Fetal reactivity is a very important CTG characteristic used to diagnose fetal distress, but its interpretation is still uncertain. The aim of this study is to characterise fetal reactivity proposing new fetal HRV frequency parameters to support a more exhaustive CTG analysis.

Linear and nonlinear parameters for the analysis of fetal heart rate signal from cardiotocographic recordings

Antepartum fetal monitoring based on the classical cardiotocography (CTG) is a noninvasive and simple tool for checking fetal status. Its introduction in the clinical routine limited the occurrence of fetal problems leading to a reduction of the precocious child mortality. Nevertheless, very poor indications on fetal pathologies can be inferred from the even automatic CTG analysis methods, which are actually employed. The feeling is that fetal heart rate (FHR) signals and uterine contractions carry much more information on fetal state than is usually extracted by classical analysis methods. In particular, FHR signal contains indications about the neural development of the fetus. However, the methods actually adopted for judging a CTG trace as "abnormal" give weak predictive indications about fetal dangers. We propose a new methodological approach for the CTG monitoring, based on a multiparametric FHR analysis, which includes spectral parameters from autoregressive models and nonlinear algorithms (approximate entropy). This preliminary study considers 14 normal fetuses, eight cases of gestational (maternal) diabetes, and 13 intrauterine growth retarded fetuses. A comparison with the traditional time domain analysis is also included. This paper shows that the proposed new parameters are able to separate normal from pathological fetuses. Results constitute the first step for realizing a new clinical classification system for the early diagnosis of most common fetal pathologies.

Empirical evaluation of a hybrid intelligent monitoring system using different measures of effectiveness

The validation of a software product is a fundamental part of its development, and focuses on an analysis of whether the software correctly resolves the problems it was designed to tackle. Traditional approaches to validation are based on a comparison of results with what is called a gold standard. Nevertheless, in certain domains, it is not always easy or even possible to establish such a standard. This is the case of intelligent systems that endeavour to simulate or emulate a model of expert behaviour. This article describes the validation of the intelligent system computer-aided foetal evaluator (CAFE), developed for intelligent monitoring of the antenatal condition based on data from the non-stress test (NST), and how this validation was accomplished through a methodology designed to resolve the problem of the validation of intelligent systems. System performance was compared to that of three obstetricians using 3450 min of cardiotocographic (CTG) records corresponding to 53 different patients. From these records different parameters were extracted and interpreted, and thus, the validation was carried out on a parameter-by-parameter basis using measurement techniques such as percentage agreement, the Kappa statistic or cluster analysis. Results showed that the system's agreement with the experts is, in general, similar to agreement between the experts themselves which, in turn, permits our system to be considered at least as skillful as our experts. Throughout our article, the results obtained are commented on with a view to demonstrating how the utilisation of different measures of the level of agreement existing between system and experts can assist not only in assessing the aptness of a system, but also in highlighting its weaknesses. This kind of assessment means that the system can be fine-tuned repeatedly to the point where the expected results are obtained.

Denoising of the fetal heart rate signal with non-linear filtering of the wavelet transform maxima

The fetal heart rate (FHR) signal provides valuable information for fetal development and well-being. However, the FHR traces derived from present-day ultrasound cardiotocographs are not of the desired quality. The paper applies the wavelet transform (WT) in order to denoise effectively the FHR signal. The denoising procedure analyses the evolution of the WT maxima across scales. The singularities of the signal create wavelet maxima with different properties from those of the induced noise. Since it is difficult to formulate precise rules that distinguish between the wavelet maxima of the FHR signal from those of the noise we have trained a neural network for this classification task. The neural network draws out successfully the noise induced wavelet maxima. An improved FHR signal can be obtained from the coarser wavelet approximation signal component and the filtered wavelet maxima by means of the inverse dyadic wavelet transform. Also, feature extraction and processing algorithms can be defined on the denoised wavelet coefficients (instead of on the original signal).

Automated analysis of antepartum fetal heart rate in relation to fetal rest-activity states: a longitudinal study of uncomplicated pregnancies using the Sonicaid System 8000

OBJECTIVES

To learn which fetal heart rate (FHR) parameters change with gestational age and to demonstrate the relation with fetal rest-activity states.

STUDY DESIGN

FHR and fetal movements were recorded in 12 uncomplicated pregnancies from 26 weeks gestational age onwards. Seventy-two FHR recordings of 60 min duration were analysed by a computer (Sonicaid System 8000). Statistical analysis of complete 60 min recordings and selective periods of rest and activity comprised Spearman's rank correlation test, regression analysis and Wilcoxon's signed-rank test.

RESULTS

The time needed to meet the system's criteria of normality decreased with gestational age. The incidence of accelerations (ACC), overall FHR variation (VAR) and variation during 'episodes of high variation' (VEHV) increased with gestational age in the total population, but statistical significance of these relations could only be demonstrated in a minority of individual fetuses. Most FHR parameters differed significantly for periods of fetal rest and activity. No FHR parameters showed a relation with gestational age during periods of rest.

CONCLUSIONS

The increase of ACC, VAR and VEHV with gestational age is primarily due to an increase during fetal activity. The considerable variation within and between fetuses, however, can only be partly explained by fetal rest-activity states.

Developments in CTG analysis

FHR monitoring has been the subject of many debates. The technique, in itself, can be considered to be accurate and reliable both in the antenatal period, when using the Doppler signal in combination with autocorrelation techniques, and during the intrapartum period, in particular when the FHR signal can be obtained from a fetal ECG electrode placed on the presenting part. The major problems with FHR monitoring relate to the reading and interpretation of the CTG tracings. Since the FHR pattern is primarily an expression of the activity of the control by the central and peripheral nervous system over cardiovascular haemodynamics, it is possibly too indirect a signal. In other specialities such as neonatology, anaesthesiology and cardiology, monitoring and graphic display of heart rate patterns have not gained wide acceptance among clinicians. Digitized archiving, numerical analysis and even more advanced techniques, as described in this chapter, have primarily found a place in obstetrics. This can be easily explained, since the obstetrician is fully dependent on indirectly collected information regarding the fetal condition, such as (a) movements experienced by the mother, observed with ultrasound or recorded with kinetocardiotocography (Schmidt, 1994), (b) perfusion of various vessels, as assessed by Doppler velocimetry, (c) the amount of amniotic fluid or (d) changes reflected in the condition of the mother, such as the development of gestation-induced hypertension and (e) the easily, continuously obtainable FHR signal. It is of particular comfort to the obstetrician that a normal FHR tracing reliably predicts the birth of the infant in a good condition, which makes cardiotocography so attractive for widespread application. However, in the intrapartum period, many traces cannot fulfil the criteria of normality, especially in the second stage. In this respect, cardiotocography remains primarily a screening and not so much a diagnostic method. As long as continuous monitoring of fetal acid-base balance has not been extensively tested in clinical practice, microblood sampling of the fetal presenting part (Saling, 1994) is a useful adjunct. The problem with non-normal tracings is that their significance is very often unclear. They may indicate serious fetal distress, finally resulting in preventable destruction of critical areas in the fetal brain and damage to various organs; or, on the contrary, they may indicate temporary changes in cardiovascular control as a reaction to the intermittent effects on fetal haemodynamics of, for example, uterine contractions, whether or not in combination with partial or complete compression of umbilical cord vessels or the vessels on the chorionic plate (van Geijn, 1994). Many factors influence the FHR and its variability, which further complicates the interpretation of FHR patterns; some have been discussed here in some detail. Undoubtedly, there is a need for quantitative and objective FHR analysis, as long as it does not lead to erroneous results. Close collaboration between engineers and clinicians is a prerequisite for further advances in this field. Decision support systems certainly have a future but only if they are able to take into account a large set of clinical data and can combine it with data obtained from FHR signals and other parameters referring to the fetal condition, such as fetal growth, Doppler velocimetry, amniotic fluid volume and biochemical and biophysical data obtained from the mother. Basic technical concepts inherent in computerized CTG analysis, such as sampling rate (Chang et al, 1995), signal loss, artefact detection (van Geijn et al, 1980), further processing of intervals, archiving in digitized format and monitor display, should receive considerable attention. There is still a long way to go until decision support systems find their way into obstetric practice. Further developments can only be achieved thanks to efforts of many basic and clinical researchers, wo

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

OBJECTIVE

To assess the capability of a computer software interpretation program, using intrapartum fetal heart rate and intrauterine pressure as recorded in a cardiotocogram to predict fetal acidosis at birth.

DESIGN AND SUBJECTS

A retrospective analysis of digitised fetal heart rate and uterine activity values obtained from 73 high risk women in labour.

SETTING

Two university teaching hospitals.

METHODS

A computer software program was constructed to analyse the digitised data and predict acidosis. The results of the analysis were compared with actual umbilical arterial blood pH and base excess at delivery.

RESULTS

The software cardiotocogram interpreter was able to predict a pH of less than 7.15 with an accuracy of 77%, a sensitivity of 88% and specificity of 75% in this set of data. It was able to predict a base excess of less than -8 mmol/l with an accuracy of 81%, a sensitivity of 76% and specificity of 82%.

CONCLUSIONS

A computerised method of analysing fetal heart rate and uterine activity using a simple algorithm has demonstrated a capability to predict fetal acidosis at the time of delivery. Further research in this area is warranted.

Computerised fetal heart rate analysis in labour--effect of sampling rate

OBJECTIVE

To study the effect of sampling rate on the accuracy of fetal heart rate recording in labour.

DESIGN

Prospective observational study.

METHODS AND SUBJECTS

The fetal heart rate was obtained from 153 fetuses in labour. The heart rate data was sampled at a rate of 0.5 Hz (every 2 s) and the means and standard deviations of 5-min segments compared against the mean and standard deviation for all beats in the same time interval.

RESULTS

There was a highly significant correlation between the means (r = 0.994, p < 0.001) and standard deviations (r = 0.957, p < 0.001) of FHR sampled on successive beats compared with 2 s sampling.

CONCLUSION

Two second sampling of the fetal heart rate in labour will allow a highly complex analytical algorithm to process the signal in near real time for objective analysis.

Interobserver variation in the assessment of fetal heart rate recordings

Electronic fetal heart rate monitoring (EFM) has not fulfilled its expectations. To improve its validity various attempts were made to standardize terminology and assessment of fetal heart rate (FHR) recordings. In a multinational study, 21 experienced obstetricians were asked to segment and classify FHR patterns, recorded in 13 obstetric cases. In addition, the referees were asked to give their interpretation of the FHR pattern, to assess the fetal condition and to propose obstetric management. The kappa statistic showed fair agreement among the obstetricians for the classification of accelerations, baseline segments and decelerations. Poor agreement was found when the referees had to classify baseline variability or the type of deceleration. Also, the clinical assessment of fetal condition and proposals for obstetric management showed poor agreement among the referees. We conclude there is still a lack of unequivocal terminology and definitions in the assessment of FHR recordings.

Fetal heart rate patterns and chronic exposure to antiepileptic drugs

Fetal heart rate (FHR) characteristics of fetuses exposed and not exposed to antiepileptic drugs (AEDs) were studied. FHR is considered to reflect central nervous system (CNS) integrity. Three intervals during pregnancy were investigated: 20, 32, and 38 weeks. At 32 and 38 weeks, FHR was studied in relation to quiet (C1F) and active (C2F) sleep periods. For each tracing, a baseline was determined and accelerations and decelerations were identified. To assess FHR variability, the long-term irregularity, interval difference and absolute beat-to-beat indexes, and the bandwidth were calculated for 30-s intervals between accelerations and decelerations. No marked differences were noted between study and control groups concerning basal FHR and the occurrence of accelerations. For FHR derived from the fetal ECG, all indexes of FHR variability and the bandwidth were lower for the study group as compared with the control group, although the differences did not reach statistical significance. Our study shows that chronic prenatal exposure to AEDs does not seriously interfere with modulation of fetal heart rhythm by the CNS.

Quantitated fetal heart rhythm at 20, 32 and 38 weeks of gestation and dependence on rest-activity patterns

Quantitative parameters of fetal heart rate (FHR) were automatically analysed at 20, 32 and 38 weeks of pregnancy. FHR was obtained both by the fetal ECG method and by wide range Doppler ultrasound with autocorrelation. At 32 and 38 weeks, FHR was studied in relation to fetal rest-activity according to the fetal behavioural state concept (coincidence 1F and 2F). Basal fetal heart rate was significantly higher at 20 weeks of gestation than at 32 and 38 weeks. The number of accelerations increased significantly from 20 weeks to 32 and 38 weeks for C2F periods. Parameters of FHR variability, i.e. ID, ABB, LTI indices and bandwidth, were higher during periods C2F compared to periods C1F. Lowest values of all four parameters were found at 20 weeks gestation. The ID index, which is a measure of short-term variability increased significantly between 32 to 38 (C2F). The absolute values of ID, ABB and LTI were lower for ultrasound recordings than for the fetal ECG.

Automated analysis of near-term antepartum fetal heart rate in relation to fetal behavioral states: the Sonicaid System 8000

Fetal heart rate and fetal movements were recorded in 16 uncomplicated near-term pregnancies. The recordings were used to evaluate a system for automated fetal heart rate analysis (Sonicaid System 8000). Fetal rest-activity patterns were considered in the analysis. The mean duration of C2F periods "active sleep," 33 minutes) was significantly greater than that of C1F periods ("quiet sleep," 19 minutes) (p less than 0.001). The incidence of accelerations and decelerations and the overall fetal heart rate variations were greater during C2F than during C1F (p less than 0.001). In 11 of 16 C1F periods, the system classified the fetal heart rate variation as "questionable" or "abnormal." Episodes of high variation were identified in only 3 of 16 C1F periods, but they were found in all 18 C2F periods. Episodes of low variation were identified in 14 of the 16 C1F periods but were not found in any C2F periods. During C2F periods, the system's criteria of normality were met in all cases but one; they were not met during any of the C1F periods. Thus the diagnosis of fetal distress should not be based merely on the absence of accelerations, low fetal heart rate variation, or absence of episodes of high variation in recordings with a duration of less than 45 minutes.

Duration, amplitude and shape of accelerations in relation to fetal body movements in behavioral state 2F

The relationship between fetal movements and fetal heart rate accelerations was analyzed in 34 healthy near term fetuses. Periods of coincidence 2F (C2F) with a mean duration of 34 +/- 6 minutes per fetus were selected, with a total of 463 accelerations. Nineteen percent of single body movements and 71% of compilations of movements were accompanied by an acceleration. The minimal duration of single movements associated with accelerations was 4-5 seconds. Movements with associated accelerations differed significantly in duration from movements without accelerations. The duration of accelerations was strongly correlated with the duration of movements. The amplitude of accelerations was not clearly correlated with the duration of movements, but depended on the type of movement. The shape of accelerations appeared to be dependent on the timing of the various fetal movements. In 77%, the number of notches in the accelerations was equal to the number of pauses in the movement complications. A discrepancy between notches in accelerations and pauses in movements could be explained in the majority of cases when the timing of the various movements in relation to one another was considered, or by the presence of fetal mouth movements.

Spectral analysis of fetal heart rhythm in relation to fetal regular mouthing

Fetal heart rate variation during fetal regular mouthing in behavioural state 1F was investigated applying spectral analysis. Periods with and without fetal regular mouthing movements were compared. The power spectrum of the periods with regular mouthing movements showed a peak at the frequency of the clusters of mouthing movements which was absent in the power spectrum of the corresponding periods without movements. The oscillations in the fetal heart rate associated with this peak in the power spectrum were detectable both in the heart rate tracings obtained from the abdominal electrocardiogram and those recorded by means of wide range Doppler ultrasound.

Computer analysis of antepartum fetal heart rate: 1. Baseline determination

A consequent and reproducible determination of baseline is an essential prerequisite for objective interpretation of fetal heart rate. A fully automated off-line method of baseline determination has been developed and tested on 50 normal antepartum fetal heart rate recordings of two hours duration. The method is constructed around two functional units, a digital filter and a trim function, which interact in an iterative process. The results were evaluated in comparison with automated baseline determination according to Dawes and coworkers. A panel of 3 experts agreed that in 14 of the 50 recordings (28%), the new developed procedure resulted in a substantially better baseline fit. In the remaining 34 recordings (72%), baseline fit from both methods was judged as equivalent. The described procedure of baseline determination provides a solid base for automated detection of accelerations and decelerations in fetal heart rate recordings. It enables the study of the relation between the fetal heart rate pattern and fetal movements. Finally, it provides an objective tool for analysis of variables within the fetal heart rate with the highest predictive value with respect to fetal outcome.