Real-time automated detection of clonic seizures in newborns

doi:10.1016/j.clinph.2013.12.119

Clinical Neurophysiology

Volume 125, Issue 8, August 2014, Pages 1533-1540

https://doi.org/10.1016/j.clinph.2013.12.119 Get rights and content

Highlights

•
In an attempt to overcome challenges in neonatal seizure recognition, automated detection systems have been developed.
•
Here we describe a new algorithm for real-time, low-cost clonic neonatal seizures detection based on differential average luminance signal analysis.
•
Encouraging sensitivity, specificity and discriminatory power suggest its wider use as a screening tool.

Abstract

Objective

The aim of this study is to apply a real-time algorithm for clonic neonatal seizures detection, based on a low complexity image processing approach extracting the differential average luminance from videotaped body movements.

Methods

23 video-EEGs from 12 patients containing 78 electrographically confirmed neonatal seizures of clonic type were reviewed and all movements were divided into noise, random movements, clonic seizures or other seizure types. Six video-EEGs from 5 newborns without seizures were also reviewed. Videos were then separately analyzed using either single, double or triple windows (these latter with 50% overlap) each of a 10 s duration.

Results

With a decision threshold set at 0.5, we obtained a sensitivity of 71% (corresponding specificity: 69%) with double-window processing for clonic seizures diagnosis. The discriminatory power, indicated by the Area Under the Curve (AUC), is higher with two interlaced windows (AUC = 0.796) than with single (AUC = 0.788) or triple-window (AUC = 0.728). Among subjects without neonatal seizures, our algorithm showed a specificity of 91% with double-window processing.

Conclusions

Our algorithm reliably detects neonatal clonic seizures and differentiates them from either noise, random movements and other seizure types.

Significance

It could represent a low-cost, low complexity, real-time automated screening tool for clonic neonatal seizures.

Introduction

Seizures are the most common symptom of acute neurological disease in newborns (Volpe, 2001). The incidence rate, as reported in population-based studies, corresponds to 2.6 per 1000 live births, increasing to 11.1‰ for preterm neonates and to 13.5‰ for infants with a birth weight lower than 2500 g (Ronen et al., 1999). Therefore, this may represent a common neurological sign in the neonatal intensive care unit (NICU) and, furthermore, could carry an increased risk of long-term morbidity (Mizrahi and Clancy, 2000).

Thus, neonatal seizures have to be promptly and accurately recognized in order to establish timely treatments. Although the traditional method of diagnosis is based on Video-ElectroEncephaloGraphic (v-EEG) monitoring, EEG interpretation is a time-consuming technique requiring specialised skills, not always readily available in the neonatal intensive care setting (Ntonfo et al., 2012). Therefore, automatic and real-time diagnostic equipment able to reliably recognize neonatal seizures would be of significant value (Kilbride et al., 2009, Shah et al., 2012, Alegre and Urrestarazu, 2011). Automatic detection of seizures by analysing EEG abnormalities has been considered (Deburchgraeve et al., 2008, Cherian et al., 2011). Alternatively, the movements of the newborn’s body could be acquired through a video camera and the corresponding video signal properly processed, with the aim to detect the newborn’s “unusual” movements (Ntonfo et al., 2012). The acquisition of the motion strength (through image processing techniques) has been proposed as an expedient to detect the presence of neonatal seizures (Kilbride et al., 2009). To achieve this objective, clonic seizures were detected by analysing relevant motion trajectory features for gesture recognition (Karayiannis et al., 2006a, Karayiannis et al., 2006b, Karayiannis et al., 2006c).

In this paper, we rely on the low complexity image processing-based approach to the detection of clonic neonatal seizures proposed in (Ntonfo et al., 2012). This method consists of the extraction of an average differential luminance signal from acquired videos, where the average is carried out over all pixels of the difference between consecutive frames. Therefore, as periodic body movements lead to periodic average luminance signals, seizure detection reduces to periodicity detection. This low-complexity algorithm naturally leads to the implementation of low-cost camera-based diagnostic apparels to assist clinical practice (Kouamou et al., 2011, Ntonfo et al., 2012).

The aim of this study is to apply in a clinical setting a new real-time algorithm (Ntonfo et al., 2012) for clonic neonatal seizures detection, based on a low complexity image processing approach extracting the differential average luminance from videotaped body movements.

Section snippets

Methods

This study was conducted on video-EEG recordings collected in the neonatal seizures database elaborated by the Child Neuropsychiatry Unit of the Neuroscience Department at Parma University. This database collects all neonatal seizures of newborns consecutively admitted to the NICU of Parma University-Hospital between June 2001 and August 2012. In our Unit all newborns at high risk of seizures, on the basis of predisposing factors (such as birth asphyxia, sepsis, meningitis, metabolic disorders,

Results

The considered approach, outlined above, has been tuned in order to guarantee good sensitivity and specificity values. The ROC curves obtained are shown below (Fig. 4): they express the varying ratio between TPs (sensitivity) and false alarms (1-specificity) for single, double or triple window, respectively. In particular, an optimized value of the decision threshold has been determined by means of ROC curves, more specifically, the best value of the decision threshold has been selected as the

Discussion

Automatic detection systems have been developed to overcome limitations in neonatal seizure identification mainly linked to the difficulties in clinical recognition and the need for highly specialised expertise to correctly interpret v-EEG. Alternative strategies already used in clinical practice, for example amplitude-integrated EEG, have proven very useful but unable to completely substitute conventional v-EEG, due to various drawbacks, such as the use of a limited set of electrodes, the

Disclosure

All the authors report no conflicts of interest or financial disclosures. The manuscript does not report results of a clinical trial. This study is not industry-sponsored.

References (23)

M. Alegre et al.
Neonatal automated seizure detection: going ahead into clinical use
Clin Neurophysiol
(2011)
P.J. Cherian et al.
Validation of a new automated neonatal seizure detection system: a clinician’s perspective
Clin Neurophysiol
(2011)
W. Deburchgraeve et al.
Automated neonatal seizure detection mimicking a human observer reading EEG
Clin Neurophysiol
(2008)
S. Faul et al.
An evaluation of automated neonatal seizure detection methods
Clin Neurophysiol
(2005)
B.R. Greene et al.
Combination of EEG and ECG for improved automatic neonatal seizure detection
Clin Neurophysiol
(2007)
N.B. Karayiannis et al.
Automated detection of videotaped neonatal seizures based on motion segmentation methods
Clin Neurophysiol
(2006)
G. Ronen et al.
The epidemiology of clinical neonatal seizures in Newfoundland: a population-based study
J Pediatr
(1999)
D. Selton et al.
EEG and ischemic stroke in full-term newborns
Neurophysiol Clin
(2003)
N.B. Karayiannis
Computerized motion analysis of videotaped neonatal seizures of epileptic origin
Epilepsia
(2005)
N.B. Karayiannis et al.
Quantifying motion in video recordings of neonatal seizures by regularized optical flow methods
IEEE Trans Image Process
(2005)

N.B. Karayiannis et al.

Automated detection of videotaped neonatal seizures of epileptic origin

Epilepsia

(2006)

Cited by (24)

Continuous sensing and quantification of body motion in infants: A systematic review
2023, Heliyon
Abnormal body motion in infants may be associated with neurodevelopmental delay or critical illness. In contrast to continuous patient monitoring of the basic vitals, the body motion of infants is only determined by discrete periodic clinical observations of caregivers, leaving the infants unattended for observation for a longer time. One step to fill this gap is to introduce and compare different sensing technologies that are suitable for continuous infant body motion quantification. Therefore, we conducted this systematic review for infant body motion quantification based on the PRISMA method (Preferred Reporting Items for Systematic Reviews and Meta-Analyses). In this systematic review, we introduce and compare several sensing technologies with motion quantification in different clinical applications. We discuss the pros and cons of each sensing technology for motion quantification. Additionally, we highlight the clinical value and prospects of infant motion monitoring. Finally, we provide suggestions with specific needs in clinical practice, which can be referred by clinical users for their implementation. Our findings suggest that motion quantification can improve the performance of vital sign monitoring, and can provide clinical value to the diagnosis of complications in infants.
RGB-D scene analysis in the NICU
2021, Computers in Biology and Medicine
Citation Excerpt :
This study aims to provide analogous information from the neonatal intensive care unit (NICU) environment to assist clinical staff in patient care and monitoring. Previous video-based studies in the NICU have focused on non-contact vital sign monitoring [1,2], behavioral and pain assessment from facial expression [3–5], and clonic seizure detection [6,7]. Other studies have integrated depth-sensing cameras to capture additional perspectives pertinent to monitoring algorithms [8–10].
Continuity of care is achieved in the neonatal intensive care unit (NICU) through careful documentation of all events of clinical significance, including clinical interventions and routine care events (e.g., feeding, diaper change, weighing, etc.). As a step towards automating this documentation process, we propose a scene recognition algorithm that can automatically identify key features in a single image of the patient environment, paired with a rule-based sentence generator to caption the scene. Color and depth video were obtained from 29 newborn patients from the Children's Hospital of Eastern Ontario (CHEO) using an Intel RealSense SR300 RGB-D camera and manual bedside event annotation. Image processing techniques are implemented to classify two lighting conditions: brightness level and phototherapy. A deep neural network is developed for three image classification tasks: on-going intervention, bed occupancy, and patient coverage. Transfer learning is leveraged in the feature extraction layers, such that weights learned from a generic data-rich task are applied to the clinical domain where data collection is complex and costly. Different depth fusion techniques are implemented and compared among classification tasks, where the depth and color data are fused as an RGB-D image (image fusion) or separately at various layers in the network (network fusion). Promising results were obtained with >84% sensitivity and >73% F1 measure across all context variables despite the large class imbalance. RGBD-based models are shown to outperform RGB models on most tasks. In general, a 4-channel image fusion and network fusion at the 11th layer of the VGG-16 architecture were preferred. Ultimately, achieving complete scene understanding through multimodal computer vision could form the basis for a semi-automated charting system to assist clinical staff.
Seizures in the neonate: A review of etiologies and outcomes
2021, Seizure
Citation Excerpt :
aEEG is very effective in the detection of prolonged seizures and status epilepticus [33], but may miss brief and focal seizures as well as those associated with low-amplitude ictal EEG patterns that constitute a significant proportion of neonatal seizures, particularly in preterms [34–36]. In addition, recently developed seizure detection algorithms deriving from EEG and EKG data [37], from EEG data alone [38–40], or from movement analysis [29,41–43], have the potential to overcome the hurdles of continuous EEG monitoring interpretation, and facilitate as well as expedite the diagnosis of neonatal seizures by providing a real-time decision support. Overall, automated seizure detection is promising, but still requires validation for use in everyday clinical practice.
Neonatal seizures occur in their majority in close temporal relation to an acute brain injury or systemic insult, and are accordingly defined as acute symptomatic or provoked seizures. However less frequently, unprovoked seizures may also present in the neonatal period as secondary to structural brain abnormalities, thus corresponding to structural epilepsies, or to genetic conditions, thus corresponding to genetic epilepsies. Unprovoked neonatal seizures should be thus considered as the clinical manifestation of early onset structural or genetic epilepsies that often have the characteristics of early onset epileptic encephalopathies. In this review, we address the conundrum of neonatal seizures including acute symptomatic, remote symptomatic, provoked, and unprovoked seizures, evolving to post-neonatal epilepsies, and neonatal onset epilepsies. The different clinical scenarios involving neonatal seizures, each with their distinct post-neonatal evolution are presented. The structural and functional impact of neonatal seizures on brain development and the concept of secondary epileptogenesis, with or without a following latent period after the acute seizures, are addressed. Finally, we underline the need for an early differential diagnosis between an acute symptomatic seizure and an unprovoked seizure, since it is associated with fundamental differences in clinical evolution. These are crucial aspects for neonatal management, counselling and prognostication. In view of the above aspects, we provide an outlook on future strategies and potential lines of research in this field.
Design and implementation of a multi-sensor newborn EEG seizure and background model with inter-channel field characterization
2019, Digital Signal Processing: A Review Journal
Citation Excerpt :
They represent a major risk factor for morbidity and death, and may contribute to various detrimental neurological outcomes [7,17,18]. Seizures occur in 1 to 5 per 1000 live births, increasing to 11.1% for preterm newborns and to 13.5% for infants with a birth weight less than 2.5 kilograms [2,19,20]. In addition, newborns with seizures have been reported to be 55 to 70 times more likely to develop severe cerebral palsy [13].
This paper presents a novel multi-sensor non-stationary EEG model; it is obtained by combining state of the art mono-sensor newborn EEG simulators, a multilayer newborn head model comprised of four homogeneous concentric spheres, a multi-sensor propagation scheme based on array processing and optical dispersion to calculate inter-channel attenuation and delay, and lastly, a multi-variable optimization paradigm using particle swarm optimization and Monte-Carlo simulations to validate the model for optimal conditions. Multi-sensor EEG of 7 newborns, comprised of seizure and background epochs, are analyzed using time-space, time-frequency, power maps and multi-sensor causality techniques. The outcomes of these methods are validated by medical insights and serve as a backbone for any assumptions and as performance benchmarks for the model to be evaluated against. The results obtained with the developed model show 85.7% averaged time-frequency correlation (which is the selected measure for similarity with real EEG) with 5.9% standard deviation, and the averaged error obtained is 34.6% with 8% standard deviation. The resulting performances indicate that the proposed model provides a suitable matching fit with real EEG in terms of their probability density function, inter-sensor attenuation and translation, and multi-sensor causality. They also demonstrate the model flexibility to generate new unseen samples by utilizing user-defined parameters, making it suitable for other relevant applications.
Monitoring infants by automatic video processing: A unified approach to motion analysis
2017, Computers in Biology and Medicine
A unified approach to contact-less and low-cost video processing for automatic detection of neonatal diseases characterized by specific movement patterns is presented. This disease category includes neonatal clonic seizures and apneas. Both disorders are characterized by the presence or absence, respectively, of periodic movements of parts of the body—e.g., the limbs in case of clonic seizures and the chest/abdomen in case of apneas. Therefore, one can analyze the data obtained from multiple video sensors placed around a patient, extracting relevant motion signals and estimating, using the Maximum Likelihood (ML) criterion, their possible periodicity. This approach is very versatile and allows to investigate various scenarios, including: a single Red, Green and Blue (RGB) camera, an RGB-depth sensor or a network of a few RGB cameras. Data fusion principles are considered to aggregate the signals from multiple sensors. In the case of apneas, since breathing movements are subtle, the video can be pre-processed by a recently proposed algorithm which is able to emphasize small movements. The performance of the proposed contact-less detection algorithms is assessed, considering real video recordings of newborns, in terms of sensitivity, specificity, and Receiver Operating Characteristic (ROC) curves, with respect to medical gold standard devices. The obtained results show that a video processing-based system can effectively detect the considered specific diseases, with increasing performance for increasing number of sensors.
Seizure reporting technologies for epilepsy treatment: A review of clinical information needs and supporting technologies
2015, Seizure
Citation Excerpt :
In our review, inertial seizure detection systems [2,15,30,57] achieved higher performance than embedded mattress devices [7,63] and multimodal devices [49]. Inertial devices also tend to support daytime use, while mattress and video systems are often limited nighttime use within bedrooms [48,63]. More accurate seizure counts could better inform AED treatment.
This review surveys current seizure detection and classification technologies as they relate to aiding clinical decision-making during epilepsy treatment. Interviews and data collected from neurologists and a literature review highlighted a strong need for better distinguishing between patients exhibiting generalized and partial seizure types as well as achieving more accurate seizure counts. This information is critical for enabling neurologists to select the correct class of antiepileptic drugs (AED) for their patients and evaluating AED efficiency during long-term treatment. In our questionnaire, 100% of neurologists reported they would like to have video from patients prior to selecting an AED during an initial consultation. Presently, only 30% have access to video. In our technology review we identified that only a subset of available technologies surpassed patient self-reporting performance due to high false positive rates. Inertial seizure detection devices coupled with video capture for recording seizures at night could stand to address collecting seizure counts that are more accurate than current patient self-reporting during day and night time use.

View all citing articles on Scopus

^☆: GM Kouamou Ntonfo completed the statistical analysis.

View full text

Real-time automated detection of clonic seizures in newborns☆

Highlights

Abstract

Objective

Methods

Results

Conclusions

Significance

Introduction

Section snippets

Methods

Results

Discussion

Disclosure

Clin Neurophysiol

Clin Neurophysiol

Clin Neurophysiol

Clin Neurophysiol

Clin Neurophysiol

Clin Neurophysiol

J Pediatr

Neurophysiol Clin

Computerized motion analysis of videotaped neonatal seizures of epileptic origin

Epilepsia

Quantifying motion in video recordings of neonatal seizures by regularized optical flow methods

IEEE Trans Image Process

Automated detection of videotaped neonatal seizures of epileptic origin

Epilepsia