Automatic evaluation of intrapartum fetal heart rate recordings: a comprehensive analysis of useful features

Computerised Cardiotocography Analysis for the Automated Detection of Fetal Compromise during Labour: A Review

The measurement and analysis of fetal heart rate (FHR) and uterine contraction (UC) patterns, known as cardiotocography (CTG), is a key technology for detecting fetal compromise during labour. This technology is commonly used by clinicians to make decisions on the mode of delivery to minimise adverse outcomes. A range of computerised CTG analysis techniques have been proposed to overcome the limitations of manual clinician interpretation. While these automated techniques can potentially improve patient outcomes, their adoption into clinical practice remains limited. This review provides an overview of current FHR and UC monitoring technologies, public and private CTG datasets, pre-processing steps, and classification algorithms used in automated approaches for fetal compromise detection. It aims to highlight challenges inhibiting the translation of automated CTG analysis methods from research to clinical application and provide recommendations to overcome them.

The Effect of Fetal Heart Rate Segment Selection on Deep Learning Models for Fetal Compromise Detection

Monitoring the fetal heart rate (FHR) is common practice in obstetric care to assess the risk of fetal compromise. Unfortunately, human interpretation of FHR recordings is subject to inter-observer variability with high false positive rates. To improve the performance of fetal compromise detection, deep learning methods have been proposed to automatically interpret FHR recordings. However, existing deep learning methods typically analyse a fixed-length segment of the FHR recording after removing signal gaps, where the influence of this segment selection process has not been comprehensively assessed. In this work, we develop a novel input length invariant deep learning model to determine the effect of FHR segment selection for detecting fetal compromise. Using this model, we perform five times repeated five-fold cross-validation on an open-access database of 552 FHR recordings and assess model performance for FHR segment lengths between 15 and 60 minutes. We show that the performance after removing signal gaps improves with increasing segment length from 15 minutes (AUC = 0.50) to 60 minutes (AUC = 0.74). Additionally, we demonstrate that using FHR segments without removing signal gaps achieves superior performance across signal lengths from 15 minutes (AUC = 0.68) to 60 minutes (AUC = 0.76). These results show that future works should carefully consider FHR segment selection and that removing signal gaps might contribute to the loss of valuable information.

Artificial Intelligence and Women Researchers in the Czech Republic

Artificial intelligence as a research area has been continuously growing for several decades. Many applications were developed in various domains. Medicine and health care have attracted more intensive attention thanks to rapid technological development that has accelerated generation of large volumes of data requiring intelligent analysis and evaluation. This article illustrates, through examples of women researchers and selected AI projects in medicine, the wide spectrum of applications developed during the last fifteen years in the Czech Republic, and in particular at the Czech Technical University in Prague. Women researchers played an important and irreplaceable role since the advent of AI research in the Czech Republic. By their example, they motivated many young female students to join the community and start their research career in the AI area. They frequently participated in research projects led by the senior women researchers. The presented overview of projects illustrates the diversity of the medical area and the potential of AI methods that can be used for solving data- and knowledge-intensive problems. We briefly touch on the AI study programs. In conclusion, we point out the future challenges in AI and its applications in medicine and health care.

Non-linear Methods Predominant in Fetal Heart Rate Analysis: A Systematic Review

The analysis of fetal heart rate variability has served as a scientific and diagnostic tool to quantify cardiac activity fluctuations, being good indicators of fetal well-being. Many mathematical analyses were proposed to evaluate fetal heart rate variability. We focused on non-linear analysis based on concepts of chaos, fractality, and complexity: entropies, compression, fractal analysis, and wavelets. These methods have been successfully applied in the signal processing phase and increase knowledge about cardiovascular dynamics in healthy and pathological fetuses. This review summarizes those methods and investigates how non-linear measures are related to each paper's research objectives. Of the 388 articles obtained in the PubMed/Medline database and of the 421 articles in the Web of Science database, 270 articles were included in the review after all exclusion criteria were applied. While approximate entropy is the most used method in classification papers, in signal processing, the most used non-linear method was Daubechies wavelets. The top five primary research objectives covered by the selected papers were detection of signal processing, hypoxia, maturation or gestational age, intrauterine growth restriction, and fetal distress. This review shows that non-linear indices can be used to assess numerous prenatal conditions. However, they are not yet applied in clinical practice due to some critical concerns. Some studies show that the combination of several linear and non-linear indices would be ideal for improving the analysis of the fetus's well-being. Future studies should narrow the research question so a meta-analysis could be performed, probing the indices' performance.

Long-term prognostic counselling in people with multiple sclerosis using an online analytical processing tool

BACKGROUND

Prognostic counselling is a sensitive issue in medicine and especially so in MS due to the highly heterogeneous disease course. However, people with MS (pwMS) seek prognostic information. The web-based 'Evidence-Based Decision Support Tool in Multiple Sclerosis' (EBDiMS) uses data of 717 patients from the London/Ontario cohort to calculate personalized long-term prognostic information.

OBJECTIVE

The aim of this study was to investigate the feasibility and effect of long-term prognostic counselling in pwMS using EBDiMS.

METHODS

Ninety consecutive pwMS were provided with personalized estimations of expected time to reach Expanded Disability Status Scale (EDSS) scores of 6 and 8 and time to conversion to secondary-progressive MS. Participants gave estimates on their own putative prognosis and rated the tool's acceptability on six-step Likert-type scales.

RESULTS

Participants rated EBDiMS as highly understandable, interesting and relevant for patient-physician encounters, coping and therapy decisions. Although it provoked a certain degree of worry in some participants, 95% would recommend using the tool. Participants' own prognosis estimates did not change significantly following EBDiMS.

CONCLUSION

Long-term prognostic counselling using an online tool has been shown to be feasible in a clinical setting. EBDiMS provides pwMS with relevant, easy-to-understand, long-term prognostic information without causing relevant anxiety.

DeepFHR: intelligent prediction of fetal Acidemia using fetal heart rate signals based on convolutional neural network

Background

Fetal heart rate (FHR) monitoring is a screening tool used by obstetricians to evaluate the fetal state. Because of the complexity and non-linearity, a visual interpretation of FHR signals using common guidelines usually results in significant subjective inter-observer and intra-observer variability. Objective: Therefore, computer aided diagnosis (CAD) systems based on advanced artificial intelligence (AI) technology have recently been developed to assist obstetricians in making objective medical decisions.

Methods

In this work, we present an 8-layer deep convolutional neural network (CNN) framework to automatically predict fetal acidemia. After signal preprocessing, the input 2-dimensional (2D) images are obtained using the continuous wavelet transform (CWT), which provides a better way to observe and capture the hidden characteristic information of the FHR signals in both the time and frequency domains. Unlike the conventional machine learning (ML) approaches, this work does not require the execution of complex feature engineering, i.e., feature extraction and selection. In fact, 2D CNN model can self-learn useful features from the input data with the prerequisite of not losing informative features, representing the tremendous advantage of deep learning (DL) over ML.

Results

Based on the test open-access database (CTU-UHB), after comprehensive experimentation, we achieved better classification performance using the optimal CNN configuration compared to other state-of-the-art methods: the averaged ten-fold cross-validation of the accuracy, sensitivity, specificity, quality index defined as the geometric mean of the sensitivity and specificity, and the area under the curve yielded results of 98.34, 98.22, 94.87, 96.53 and 97.82%, respectively

Conclusions

Once the proposed CNN model is successfully trained, the corresponding CAD system can be served as an effective tool to predict fetal asphyxia objectively and accurately.

Cardiotocography and beyond: a review of one-dimensional Doppler ultrasound application in fetal monitoring

One-dimensional Doppler ultrasound (1D-DUS) provides a low-cost and simple method for acquiring a rich signal for use in cardiovascular screening. However, despite the use of 1D-DUS in cardiotocography (CTG) for decades, there are still challenges that limit the effectiveness of its users in reducing fetal and neonatal morbidities and mortalities. This is partly due to the noisy, transient, complex and nonstationary nature of the 1D-DUS signals. Current challenges also include lack of efficient signal quality metrics, insufficient signal processing techniques for extraction of fetal heart rate and other vital parameters with adequate temporal resolution, and lack of appropriate clinical decision support for CTG and Doppler interpretation. Moreover, the almost complete lack of open research in both hardware and software in this field, as well as commercial pressures to market the much more expensive and difficult to use Doppler imaging devices, has hampered innovation. This paper reviews the basics of fetal cardiac function, 1D-DUS signal generation and processing, its application in fetal monitoring and assessment of fetal development and wellbeing. It also provides recommendations for future development of signal processing and modeling approaches, to improve the application of 1D-DUS in fetal monitoring, as well as the need for annotated open databases.

Investigating pH based evaluation of fetal heart rate (FHR) recordings

Cardiotocography (CTG) is a standard tool for the assessment of fetal well-being during pregnancy and delivery. However, its interpretation is associated with high inter- and intra-observer variability. Since its introduction there have been numerous attempts to develop computerized systems assisting the evaluation of the CTG recording. Nevertheless these systems are still hardly used in a delivery ward. Two main approaches to computerized evaluation are encountered in the literature; the first one emulates existing guidelines, while the second one is more of a data-driven approach using signal processing and computational methods. The latter employs preprocessing, feature extraction/selection and a classifier that discriminates between two or more classes/conditions. These classes are often formed using the umbilical cord artery pH value measured after delivery. In this work an approach to Fetal Heart Rate (FHR) classification using pH is presented that could serve as a benchmark for reporting results on the unique open-access CTU-UHB CTG database, the largest and the only freely available database of this kind. The overall results using a very small number of features and a Least Squares Support Vector Machine (LS-SVM) classifier, are in accordance to the ones encountered in the literature and outperform the results of a baseline classification scheme proving the utility of using advanced data processing methods. Therefore the achieved results can be used as a benchmark for future research involving more informative features and/or better classification algorithms.

Sparse Support Vector Machine for Intrapartum Fetal Heart Rate Classification

Fetal heart rate (FHR) monitoring is routinely used in clinical practice to help obstetricians assess fetal health status during delivery. However, early detection of fetal acidosis that allows relevant decisions for operative delivery remains a challenging task, receiving considerable attention. This contribution promotes sparse support vector machine classification that permits to select a small number of relevant features and to achieve efficient fetal acidosis detection. A comprehensive set of features is used for FHR description, including enhanced and computerized clinical features, frequency domain, and scaling and multifractal features, all computed on a large (1288 subjects) and well-documented database. The individual performance obtained for each feature independently is discussed first. Then, it is shown that the automatic selection of a sparse subset of features achieves satisfactory classification performance (sensitivity 0.73 and specificity 0.75, outperforming clinical practice). The subset of selected features (average depth of decelerations MAD_dtrd, baseline level β₀ , and variability H) receives simple interpretation in clinical practice. Intrapartum fetal acidosis detection is improved in several respects: A comprehensive set of features combining clinical, spectral, and scale-free dynamics is used; an original multivariate classification targeting both sparse feature selection and high performance is devised; state-of-the-art performance is obtained on a much larger database than that generally studied with description of common pitfalls in supervised classification performance assessments.

Intrapartum fetal heart rate classification from trajectory in Sparse SVM feature space

Intrapartum fetal heart rate (FHR) constitutes a prominent source of information for the assessment of fetal reactions to stress events during delivery. Yet, early detection of fetal acidosis remains a challenging signal processing task. The originality of the present contribution are three-fold: multiscale representations and wavelet leader based multifractal analysis are used to quantify FHR variability ; Supervised classification is achieved by means of Sparse-SVM that aim jointly to achieve optimal detection performance and to select relevant features in a multivariate setting ; Trajectories in the feature space accounting for the evolution along time of features while labor progresses are involved in the construction of indices quantifying fetal health. The classification performance permitted by this combination of tools are quantified on a intrapartum FHR large database (≃ 1250 subjects) collected at a French academic public hospital.

Fractal Analysis and Hurst Parameter for Intrapartum Fetal Heart Rate Variability Analysis: A Versatile Alternative to Frequency Bands and LF/HF Ratio

Background

The fetal heart rate (FHR) is commonly monitored during labor to detect early fetal acidosis. FHR variability is traditionally investigated using Fourier transform, often with adult predefined frequency band powers and the corresponding LF/HF ratio. However, fetal conditions differ from adults and modify spectrum repartition along frequencies.

Aims

This study questions the arbitrariness definition and relevance of the frequency band splitting procedure, and thus of the calculation of the underlying LF/HF ratio, as efficient tools for characterizing intrapartum FHR variability.

Study Design

The last 30 minutes before delivery of the intrapartum FHR were analyzed.

Subjects

Case-control study. A total of 45 singletons divided into two groups based on umbilical cord arterial pH: the Index group with pH ≤ 7.05 (n = 15) and Control group with pH > 7.05 (n = 30).

Outcome Measures

Frequency band-based LF/HF ratio and Hurst parameter.

Results

This study shows that the intrapartum FHR is characterized by fractal temporal dynamics and promotes the Hurst parameter as a potential marker of fetal acidosis. This parameter preserves the intuition of a power frequency balance, while avoiding the frequency band splitting procedure and thus the arbitrary choice of a frequency separating bands. The study also shows that extending the frequency range covered by the adult-based bands to higher and lower frequencies permits the Hurst parameter to achieve better performance for identifying fetal acidosis.

Conclusions

The Hurst parameter provides a robust and versatile tool for quantifying FHR variability, yields better acidosis detection performance compared to the LF/HF ratio, and avoids arbitrariness in spectral band splitting and definitions.

Sampling frequency of fetal heart rate impacts the ability to predict pH and BE at birth: a retrospective multi-cohort study

Fetal heart rate (FHR) sampling rate used on the bedside is equal or less than 4 Hz. Current FHR analysis methods fail to detect incipient fetal acidemia. In a fetal sheep model of human labour we showed that FHR sampling rates near 1000 Hz are needed to detect fetal acidemia. Trans-abdominal fetal ECG (t-a fECG) sampling FHR at 900 Hz combined with a complex signals bioinformatics approach showed promise in a human cohort. Here we validate this finding in a retrospective human cohort study by comparing the performance of the same bioinformatics approach to predict pH and BE at birth in the cohorts with FHR sampled either at 4 Hz or at 900 Hz.The 4 Hz FHR recording data sets consisted of the open access intrapartum CTG data base with n = 552 subjects used to develop the predictive model and another cohort of prospectively recruited n = 11 labouring women to then validate it. 900 Hz FHR data set comprised two prospectively recruited t-a fECG cohorts of n = 60 and n = 23 subjects. Recruitment criteria were similar across the cohorts. We have determined the goodness of fit (R(2)) and root mean square error (RMSE) as the performance indicators of the model on each cohort.The clinical characteristics of all cohorts were similar (gestational age 280   ±   8 d; gender 50% male; birth body weight 3.5   ±   0.5 kg; pH and BE at birth 7.25   ±   0.1 and  -5.7   ±   3.4 mmol L(  -  1), respectively; 1' and 5' Apgar scores at birth 8.5   ±   1.4 and 9.4   ±   0.6, respectively). The 4 Hz FHR cohort rendered-for pH and BE-R(2) = 0.26 and 0.2 and RMSE = 0.087 and 3.44, respectively. This could not be confirmed in the validation cohort for neither pH nor BE prediction. The 900 Hz FHR cohort rendered-for pH and BE-R(2) = 0.9 and 0.77 and RMSE = 0.03 and 1.70, respectively, and the pH prediction was validated.In our model, lower FHR sampling rate increased the predicted error range ~3-4 fold. We show that increasing FHR sampling rate to 900 Hz improves prediction of fetal pH and BE at birth. This should improve early identification of babies at risk of brain injury.

Combining latent class analysis labeling with multiclass approach for fetal heart rate categorization

The most common approach to assess fetal well-being during delivery is monitoring of fetal heart rate and uterine contractions-the cardiotocogram (CTG). Nevertheless, 40 years since the introduction of CTG to clinical practice, its evaluation is still challenging with high inter- and intra-observer variability. Therefore the development of more objective methods has become an issue of major importance in the field. Unlike the usually proposed approaches to assign classes for classification methods that rely either on biochemical parameters (e.g. pH value) or a simple aggregation of expert judgment, this work investigates the use of an alternative labeling system using latent class analysis (LCA) along with an ordinal classification scheme. The study is performed on a well-documented open-access database, where nine expert obstetricians provided CTG annotations. The LCA is proposed here to produce more objective class labels while the ordinal classification aims to explore the natural ordering, and representation of increased severity, for obtaining the final results. The results are promising suggesting that more effort should be put into this proposed approach.

Comparison of a novel computerized analysis program and visual interpretation of cardiotocography

Objective

To compare a novel computerized analysis program with visual cardiotocography (CTG) interpretation results.

Methods

Sixty-two intrapartum CTG tracings with 20- to 30-minute sections were independently interpreted using a novel computerized analysis program, as well as the visual interpretations of eight obstetricians, to evaluate the baseline fetal heart rate (FHR), baseline FHR variability, number of accelerations, number/type of decelerations, uterine contraction (UC) frequency, and the National Institute of Child Health and Human Development (NICHD) 3-Tier FHR classification system.

Results

There was no significant difference in interobserver variation after adding the components of computerized analysis to results from the obstetricians' visual interpretations, with excellent agreement for the baseline FHR (ICC 0.91), the number of accelerations (ICC 0.85), UC frequency (ICC 0.97), and NICHD category I (kappa statistic 0.91); good agreement for baseline variability (kappa statistic 0.68), the numbers of early decelerations (ICC 0.78) and late decelerations (ICC 0.67), category II (kappa statistic 0.78), and overall categories (kappa statistic 0.80); and moderate agreement for the number of variable decelerations (ICC 0.60), and category III (kappa statistic 0.50).

Conclusions

This computerized analysis program is not inferior to visual interpretation, may improve interobserver variations, and could play a vital role in prenatal telemedicine.

Open access intrapartum CTG database

Background

Cardiotocography (CTG) is a monitoring of fetal heart rate and uterine contractions. Since 1960 it is routinely used by obstetricians to assess fetal well-being. Many attempts to introduce methods of automatic signal processing and evaluation have appeared during the last 20 years, however still no significant progress similar to that in the domain of adult heart rate variability, where open access databases are available (e.g. MIT-BIH), is visible. Based on a thorough review of the relevant publications, presented in this paper, the shortcomings of the current state are obvious. A lack of common ground for clinicians and technicians in the field hinders clinically usable progress. Our open access database of digital intrapartum cardiotocographic recordings aims to change that.

Description

The intrapartum CTG database consists in total of 552 intrapartum recordings, which were acquired between April 2010 and August 2012 at the obstetrics ward of the University Hospital in Brno, Czech Republic. All recordings were stored in electronic form in the OB TraceVue®;system. The recordings were selected from 9164 intrapartum recordings with clinical as well as technical considerations in mind. All recordings are at most 90 minutes long and start a maximum of 90 minutes before delivery. The time relation of CTG to delivery is known as well as the length of the second stage of labor which does not exceed 30 minutes. The majority of recordings (all but 46 cesarean sections) is – on purpose – from vaginal deliveries. All recordings have available biochemical markers as well as some more general clinical features. Full description of the database and reasoning behind selection of the parameters is presented in the paper.

Conclusion

A new open-access CTG database is introduced which should give the research community common ground for comparison of results on reasonably large database. We anticipate that after reading the paper, the reader will understand the context of the field from clinical and technical perspectives which will enable him/her to use the database and also understand its limitations.

Automated detection of perinatal hypoxia using time-frequency-based heart rate variability features

Perinatal hypoxia is a cause of cerebral injury in foetuses and neonates. Detection of foetal hypoxia during labour based on the pattern recognition of heart rate signals suffers from high observer variability and low specificity. We describe a new automated hypoxia detection method using time-frequency analysis of heart rate variability (HRV) signals. This approach uses features extracted from the instantaneous frequency and instantaneous amplitude of HRV signal components as well as features based on matrix decomposition of the signals' time-frequency distributions using singular value decomposition and non-negative matrix factorization. The classification between hypoxia and non-hypoxia data is performed using a support vector machine classifier. The proposed method is tested on a dataset obtained from a newborn piglet model with a controlled hypoxic insult. The chosen HRV features show strong performance compared to conventional spectral features and other existing methods of hypoxia detection with a sensitivity 93.3 %, specificity 98.3 % and accuracy 95.8 %. The high predictive value of this approach to detecting hypoxia is a substantial step towards developing a more accurate and reliable hypoxia detection method for use in human foetal monitoring.

Highly comparative fetal heart rate analysis

A database of fetal heart rate (FHR) time series measured from 7 221 patients during labor is analyzed with the aim of learning the types of features of these recordings that are informative of low cord pH. Our 'highly comparative' analysis involves extracting over 9 000 time-series analysis features from each FHR time series, including measures of autocorrelation, entropy, distribution, and various model fits. This diverse collection of features was developed in previous work [1]. We describe five features that most accurately classify a balanced training set of 59 'low pH' and 59 'normal pH' FHR recordings. We then describe five of the features with the strongest linear correlation to cord pH across the full dataset of FHR time series. The features identified in this work may be used as part of a system for guiding intervention during labor in future. This work successfully demonstrates the utility of comparing across a large, interdisciplinary literature on time-series analysis to automatically contribute new scientific results for specific biomedical signal processing challenges.

Assessment of features for automatic CTG analysis based on expert annotation

Cardiotocography (CTG) is the monitoring of fetal heart rate (FHR) and uterine contractions (TOCO) since 1960's used routinely by obstetricians to detect fetal hypoxia. The evaluation of the FHR in clinical settings is based on an evaluation of macroscopic morphological features and so far has managed to avoid adopting any achievements from the HRV research field. In this work, most of the ever-used features utilized for FHR characterization, including FIGO, HRV, nonlinear, wavelet, and time and frequency domain features, are investigated and the features are assessed based on their statistical significance in the task of distinguishing the FHR into three FIGO classes. Annotation derived from the panel of experts instead of the commonly utilized pH values was used for evaluation of the features on a large data set (552 records). We conclude the paper by presenting the best uncorrelated features and their individual rank of importance according to the meta-analysis of three different ranking methods. Number of acceleration and deceleration, interval index, as well as Lempel-Ziv complexity and Higuchi's fractal dimension are among the top five features.