Prediction of short-term health outcomes in preterm neonates from heart-rate variability and blood pressure using boosted decision trees

BACKGROUND AND OBJECTIVE

Efficient management of low blood pressure (BP) in preterm neonates remains challenging with considerable variability in clinical practice. There is currently no clear consensus on what constitutes a limit for low BP that is a risk to the preterm brain. It is argued that a personalised approach rather than a population based threshold is more appropriate. This work aims to assist healthcare professionals in assessing preterm wellbeing during episodes of low BP in order to decide when and whether hypotension treatment should be initiated. In particular, the study investigates the relationship between heart rate variability (HRV) and BP in preterm infants and its relevance to a short-term health outcome.

METHODS

The study is performed on a large clinically collected dataset of 831 h from 23 preterm infants of less than 32 weeks gestational age. The statistical predictive power of common HRV features is first assessed with respect to the outcome. A decision support system, based on boosted decision trees (XGboost), was developed to continuously estimate the probability of neonatal morbidity based on the feature vector of HRV characteristics and the mean arterial blood pressure.

RESULTS

It is shown that the predictive power of the extracted features improves when observed during episodes of hypotension. A single best HRV feature achieves an AUC of 0.87. Combining multiple HRV features extracted during hypotensive episodes with the classifier achieves an AUC of 0.97, using a leave-one-patient-out performance assessment. Finally it is shown that good performance can even be achieved using continuous HRV recordings, rather than only focusing on hypotensive events - this had the benefit of not requiring invasive BP monitoring.

CONCLUSIONS

The work presents a promising step towards the use of multimodal data in providing objective decision support for the prediction of short-term outcome in preterm infants with hypotensive episodes.

Recommendations for the design of therapeutic trials for neonatal seizures

Although seizures have a higher incidence in neonates than any other age group and are associated with significant mortality and neurodevelopmental disability, treatment is largely guided by physician preference and tradition, due to a lack of data from well-designed clinical trials. There is increasing interest in conducting trials of novel drugs to treat neonatal seizures, but the unique characteristics of this disorder and patient population require special consideration with regard to trial design. The Critical Path Institute formed a global working group of experts and key stakeholders from academia, the pharmaceutical industry, regulatory agencies, neonatal nurse associations, and patient advocacy groups to develop consensus recommendations for design of clinical trials to treat neonatal seizures. The broad expertise and perspectives of this group were invaluable in developing recommendations addressing: (1) use of neonate-specific adaptive trial designs, (2) inclusion/exclusion criteria, (3) stratification and randomization, (4) statistical analysis, (5) safety monitoring, and (6) definitions of important outcomes. The guidelines are based on available literature and expert consensus, pharmacokinetic analyses, ethical considerations, and parental concerns. These recommendations will ultimately facilitate development of a Master Protocol and design of efficient and successful drug trials to improve the treatment and outcome for this highly vulnerable population.

Coagulation Profiles Are Associated With Early Clinical Outcomes in Neonatal Encephalopathy

Introduction: Neonatal encephalopathy (NE) is associated with coagulation abnormalities. We aimed to investigate the serial alterations in coagulation profiles in term infants with NE and correlate with their clinical outcomes. This was a prospective cohort study in a tertiary referral, university-affiliated maternity hospital. Neonates exposed to perinatal asphyxia were recruited (n = 82) and 39 received therapeutic hypothermia. Infants had serial coagulation tests including platelets, prothrombin time (PT), activated partial thromboplastin time (aPTT) and fibrinogen in the first week of life. The main outcome measures included MRI brain and EEG seizures. Our results show that mortality was predicted on day 1 by decreased Fibrinogen (AUC = 0.95, p = 0.009) and by PT on day 2 with a cutoff of 22 s. An abnormal MRI was predicted by Fibrinogen on day 3 with a cut-off value of 2 g/L. For prediction of grade II/III NE, PT on day 2 of life was strongest with a cut-off value of 14 s. Only elevated APTT levels on day 1 of life were predictive of seizures (AUC = 0.65, p = 0.04). Conclusion: Coagulation parameters are strong predictors of outcomes such as abnormal NE grade, seizures, and mortality.

Modelling interactions between blood pressure and brain activity in preterm neonates

Hypotension or low blood pressure (BP) is a common problem in preterm neonates and has been associated with adverse short and long-term outcomes. Deciding when and whether to treat hypotension relies on an understanding of the relations between blood pressure and brain function. This study aims to investigate the interaction between BP and multichannel EEG in preterm infants less than 32 weeks gestational age. The mutual information is chosen to model interaction. This measure is independent of absolute values of BP and electroencephalography (EEG) power and quantifies the level of coupling between the short-term dynamics in both signals. It is shown that while adverse health conditions as measured by higher clinical risk indices for babies (CRIB II) are accompanied by consistently lower blood pressure (r=0.43), no significant correlation was observed between CRIB scores and EEG spectral power. More importantly, the chosen measure of interaction between dynamics of EEG and BP was found to be more closely related to CRIB scores (r=0.49, p-value=0.012), with higher CRIB score associated with lower levels of interaction.

Investigating the Impact of CNN Depth on Neonatal Seizure Detection Performance

This study presents a novel, deep, fully convolutional architecture which is optimized for the task of EEG-based neonatal seizure detection. Architectures of different depths were designed and tested; varying network depth impacts convolutional receptive fields and the corresponding learned feature complexity. Two deep convolutional networks are compared with a shallow SVMbased neonatal seizure detector, which relies on the extraction of hand-crafted features. On a large clinical dataset, of over 800 hours of multichannel unedited EEG, containing 1389 seizure events, the deep 11-layer architecture significantly outperforms the shallower architectures, improving the AUC90 from 82.6% to 86.8%. Combining the end-to-end deep architecture with the feature-based shallow SVM further improves the AUC90 to 87.6%. The fusion of classifiers of different depths gives greatly improved performance and reduced variability, making the combined classifier more clinically reliable.

Lost in translation? The potential psychobiotic Lactobacillus rhamnosus (JB-1) fails to modulate stress or cognitive performance in healthy male subjects

BACKGROUND

Preclinical studies have identified certain probiotics as psychobiotics - live microorganisms with a potential mental health benefit. Lactobacillus rhamnosus (JB-1) has been shown to reduce stress-related behaviour, corticosterone release and alter central expression of GABA receptors in an anxious mouse strain. However, it is unclear if this single putative psychobiotic strain has psychotropic activity in humans. Consequently, we aimed to examine if these promising preclinical findings could be translated to healthy human volunteers.

OBJECTIVES

To determine the impact of L. rhamnosus on stress-related behaviours, physiology, inflammatory response, cognitive performance and brain activity patterns in healthy male participants.

METHODS

An 8week, randomized, placebo-controlled, cross-over design was employed. Twenty-nine healthy male volunteers participated. Participants completed self-report stress measures, cognitive assessments and resting electroencephalography (EEG). Plasma IL10, IL1β, IL6, IL8 and TNFα levels and whole blood Toll-like 4 (TLR-4) agonist-induced cytokine release were determined by multiplex ELISA. Salivary cortisol was determined by ELISA and subjective stress measures were assessed before, during and after a socially evaluated cold pressor test (SECPT).

RESULTS

There was no overall effect of probiotic treatment on measures of mood, anxiety, stress or sleep quality and no significant effect of probiotic over placebo on subjective stress measures, or the HPA response to the SECPT. Visuospatial memory performance, attention switching, rapid visual information processing, emotion recognition and associated EEG measures did not show improvement over placebo. No significant anti-inflammatory effects were seen as assessed by basal and stimulated cytokine levels.

CONCLUSIONS

L. rhamnosus was not superior to placebo in modifying stress-related measures, HPA response, inflammation or cognitive performance in healthy male participants. These findings highlight the challenges associated with moving promising preclinical studies, conducted in an anxious mouse strain, to healthy human participants. Future interventional studies investigating the effect of this psychobiotic in populations with stress-related disorders are required.

Classification of hypoxic-ischemic encephalopathy using long term heart rate variability based features

Hypoxic-ischemic HI injury at the time of birth could lead to severe neurological dysfunction at an older age. Therapeutic hypothermia can be used to treat HI if severity of injury is determined within 6 hours of birth. EEG is generally used to assess the brain injury but it is neither widely recorded after birth nor is the expertise to interpret it commonly available. This study presents a novel system to classify HI injury using heart rate variability. The system makes decisions based on long-term statistical features extracted from the short-term HRV features. The preliminary results show the promising performance and robustness of the proposed method given a poor quality dataset. This tool can serve as decision support system in remote maternity units to help clinical staff to initiate hypothermia.

Pilot evaluation of the population pharmacokinetics of bumetanide in term newborn infants with seizures

Recent experimental data suggest bumetanide as a possible therapeutic option in newborn infants with seizures after birth asphyxia. Because pharmacokinetic (PK) data are lacking in this population, who very often benefit from therapeutic cooling, which can modify the PK behavior of a drug, a PK study was conducted in term infants with seizures caused by hypoxic-ischemic encephalopathy. Fourteen infants were included, 13 of them being cooled. Forty-nine blood samples were available for the determination of the plasma concentration of bumetanide. Concentration-time data were analyzed by the use of a population approach performed with Monolix Software. Bumetanide was found to follow a 2-compartment model. The mean values were 0.063 L/h for clearance, 0.28 and 0.44 L for the central and peripheral distribution volumes, respectively, and 0.59 L/h for the distribution clearance. Birth body weight explained the interindividual variability of bumetanide clearance via an allometric model. No relationship was found between bumetanide exposure and its efficacy (reduction in seizure burden) or its toxicity (hearing loss). This study describes the first PK model of bumetanide in hypothermia-treated infants with seizures.

Assessment of quality of ECG for accurate estimation of Heart Rate Variability in newborns

Heart Rate Variability has been recently used to determine the severity of Hypoxic Ischemic Encephalopathy in neonates. However, it was shown that ECG and subsequently Instantaneous Heart Rate can be heavily corrupted by artefacts which have to be manually removed. This work analyses a set of features to assess their sensitivity to normal and corrupted ECG in newborns. Specifically, the IHR signal is obtained by detecting R-Peaks using the Pan-Tompkins algorithm. Four features are extracted from both ECG and IHR signal using various temporal resolutions to discriminate normal and corrupted signal. The performance of these features in discrimination is then assessed using statistical tests.

Systematic review of neonatal seizure management strategies provides guidance on anti-epileptic treatment

There is a lack of scientific evidence to support the best management of neonatal seizures. Current strategies for neonatal seizure management were investigated by analysis of all surveys published during the time period 2000-2012. Methods for seizure diagnosis and availability of electroencephalogram (EEG), including monitoring, varied. Phenobarbital was the drug of first choice, and the use of off-label drugs and treatment times varied.

CONCLUSION

We conclude that there is an urgent need for more evidence-based studies to guide neonatal seizure management.

Clinical implementation of a neonatal seizure detection algorithm

Technologies for automated detection of neonatal seizures are gradually moving towards cot-side implementation. The aim of this paper is to present different ways to visualize the output of a neonatal seizure detection system and analyse their influence on performance in a clinical environment. Three different ways to visualize the detector output are considered: a binary output, a probabilistic trace, and a spatio-temporal colormap of seizure observability. As an alternative to visual aids, audified neonatal EEG is also considered. Additionally, a survey on the usefulness and accuracy of the presented methods has been performed among clinical personnel. The main advantages and disadvantages of the presented methods are discussed. The connection between information visualization and different methods to compute conventional metrics is established. The results of the visualization methods along with the system validation results indicate that the developed neonatal seizure detector with its current level of performance would unambiguously be of benefit to clinicians as a decision support system. The results of the survey suggest that a suitable way to visualize the output of neonatal seizure detection systems in a clinical environment is a combination of a binary output and a probabilistic trace. The main healthcare benefits of the tool are outlined. The decision support system with the chosen visualization interface is currently undergoing pre-market European multi-centre clinical investigation to support its regulatory approval and clinical adoption.

Epilepsy after neonatal seizures: literature review

INTRODUCTION

Acute neonatal seizures are the most frequent neurological complication in the neonatal intensive care units and the seizing newborns have an increased risk of long-term morbidity. However, the relationship between neonatal seizures and the development of epilepsy later in life is still unclear.

METHODS

We performed a literature review using the search terms "neonatal seizures AND outcome", "neonatal seizures AND epilepsy", "neonatal seizures AND post-neonatal epilepsy", including secondary sources of data such as reference lists of articles reviewed. From the studies in which data were available, the incidence of epilepsy was calculated by dividing the number of all subjects who developed epilepsy in the different studies considered with the number of all newborns enrolled to the studies less the number of patients lost at follow-up.

RESULTS

We found 44 studies published between 1954 and 2013, of which 4 were population-based studies and the remaining were hospital-based case series. The overall population evaluated was 4538 newborns and 17.9% developed post-neonatal epilepsy, with an onset within the first year of life in 68.5% of the patients. In 80.7%, epilepsy was associated with other neurological impairments.

CONCLUSION

Estimates on epilepsy after neonatal seizures vary widely depending on selection criteria and length of the follow-up. However, it represents a common outcome of these newborns, especially in those with severe brain injury and additional neurodevelopmental disabilities.

Neonatal EEG audification for seizure detection

Technologies for automated detection of neonatal seizures are gradually moving towards cot-side implementation. The aim of this paper is to present an alternative way to visualize the output of a neonatal seizure detection algorithm. For this purpose audified neonatal EEG is considered. The EEG is audified with the aid of the neonatal seizure detection algorithm which selects the representative channels for stereo audio image and controls the signal gain. A survey on the usefulness and accuracy of the presented audification method has been performed. The results of the audification method compare favourably to that of using amplitude integrated EEG for detection of neonatal seizures.

A novel dictionary for neonatal EEG seizure detection using atomic decomposition

The development of automated methods of electroencephalogram (EEG) seizure detection is an important problem in neonatology. This paper proposes improvements to a previously described method of seizure detection based on atomic decomposition by developing a new time-frequency (TF) dictionary that is highly coherent with the newborn EEG seizure. We compare the performance of the proposed dictionary on neonatal EEG signals with that achieved using Gabor, Fourier and wavelet dictionaries. Through the analysis of real newborn EEG data, we show first, that dictionary selection can influence the seizure detection accuracy and second, that the proposed dictionary outperforms other dictionaries by at least 10% in seizure detection accuracy and 5% improvement in the area under the Receiver Operator Characteristic curve.

Women in hospital medicine: facts, figures and personal experiences

Although females represent a high proportion of medical graduates, women are under represented at consultant level in many hospital specialties. Qualitative and quantitative analyses were undertaken which established female representation at all levels of the medical workforce in Ireland in 2011 and documented the personal experiences of a sample of female specialists. The proportions of female trainees at initial and higher specialist training levels are 765 (53%) and 656 (55%) respectively but falls to 1,685 (32%) at hospital specialist level (p < 0.0001). Significantly fewer women are found at specialist as compared to training levels in anaesthesia (p = 0.04), emergency medicine (p = 0.02), medicine (p < 0.0001), obstetrics/gynaecology (p = 0.0005), paediatrics (p = 0.006), pathology p = 0.03) and surgery (p < 0.0001). The lowest proportion of female doctors at specialist level exists in the combined surgical specialties 88 (10%); the highest is in psychiatry 380 (53%). Qualitative findings indicate that females who complete specialist training are wary of pursuing either flexible training or part time work options and experience discrimination at a number of levels. They appear to be resilient to this and tolerate it. Balancing motherhood and work commitments is the biggest challenge faced by female doctors with children and causes some to change career pathways.

Inclusion of temporal priors for automated neonatal EEG classification

The aim of this paper is to use recent advances in the clinical understanding of the temporal evolution of seizure burden in neonates with hypoxic ischemic encephalopathy to improve the performance of automated detection algorithms. Probabilistic weights are designed from temporal locations of neonatal seizure events relative to time of birth. These weights are obtained by fitting a skew-normal distribution to the temporal seizure density and introduced into the probabilistic framework of the previously developed neonatal seizure detector. The results are validated on the largest available clinical dataset, comprising 816.7 h. By exploiting these priors, the receiver operating characteristic area is increased by 23% (relative) reaching 96.74%. The number of false detections per hour is decreased from 0.45 to 0.25, while maintaining the correct detection of seizure burden at 70%.

Dynamic time warping based neonatal seizure detection system

Neonatal seizures patterns evolve with changing frequency, morphology and propagation. This study is an initial attempt to incorporate the characteristics of temporal evolution of neonatal seizures into our developed neonatal seizure detector. The previously designed SVM-based neonatal seizure detector is modified by substituting the Gaussian kernel with the Gaussian dynamic time warping kernel, to enable the SVM to classify variable length sequences of feature vectors of neonatal seizures. The preliminary results obtained compare favorably with the conventional SVM. The fusion of the two approaches is expected to improve the current state of the art neonatal seizure detection system.

Temporal evolution of seizure burden for automated neonatal EEG classification

The aim of this paper is to use recent advances in the clinical understanding of the temporal evolution of seizure burden in neonates with hypoxic ischemic encephalopathy to improve the performance of automated detection algorithms. Probabilistic weights are designed from temporal locations of neonatal seizure events relative to time of birth. These weights are obtained by fitting a skew-normal distribution to the temporal seizure density and introduced into the probabilistic framework of the previously developed neonatal seizure detector. The results are validated on the largest available clinical dataset, comprising 816.7 hours. By exploiting these priors, the ROC area is increased by 23% (relative) reaching 96.75%. The number of false detections per hour is decreased from 0.72 to 0.36, while maintaining the correct detection of seizure burden at 75%.

EEG signal description with spectral-envelope-based speech recognition features for detection of neonatal seizures

In this paper, features which are usually employed in automatic speech recognition (ASR) are used for the detection of seizures in newborn EEG. In particular, spectral envelope-based features, composed of spectral powers and their spectral derivatives are compared to the established feature set which has been previously developed for EEG analysis. The results indicate that the ASR features which model the spectral derivatives, either full-band or localized in frequency, yielded a performance improvement, in comparison to spectral-power-based features. Indeed it is shown here that they perform reasonably well in comparison with the conventional EEG feature set. The contribution of the ASR features was analyzed here using the support vector machines (SVM) recursive feature elimination technique. It is shown that the spectral derivative features consistently appear among the top-rank features. The study shows that the ASR features should be given a high priority when dealing with the description of the EEG signal.