7th Dutch Bio-Medical Engineering Conference
January 24th & 25th 2019, Egmond aan Zee, the Netherlands
13:30   Birth & Neonates
Chair: Sasa Kenjeres
15 mins
Decomposition of a Multiscale Entropy Tensor for Sleep Stage Identification in Preterm Neonates
Ofelie De Wel, Mario Lavanga, Alexander Caicedo, Katrien Jansen, Gunnar Naulaers, Sabine Van Huffel
Abstract: Preterm infants are at increased risk of neurodevelopmental impairment. Therefore, these vulnerable neonates need close monitoring of their neurological function during the first critical weeks. Sleep state organization undergoes fast development in this period and reflects the level of functional brain integrity. Hence, an automated identification of neonatal sleep stages is of great interest to evaluate electrocortical maturation. This study aimed to discriminate quiet sleep (QS) from non-quiet sleep (non-QS) using the nonlinear dynamics of the EEG signal. The analysis is performed on 97 multichannel EEGs from 26 prematurely born babies recorded at a postmenstrual age between 27 and 42 weeks. Multiscale entropy, which assesses the degree of regularity of a signal at multiple time scales, is used to quantify the complexity of each 8-channel EEG segment. The entropy values of each EEG recording are then organised in a third order tensor with modes: channels, scales and time segments. Subsequently, a rank-1 nonnegative canonical polyadic decomposition of the multiscale entropy tensor is computed, resulting in a spatial signature, a scale signature and a temporal signature. The temporal signature captures the main variation of the EEG complexity over the different time segments. After smoothing this factor with a moving average filter, k-means clustering is applied to obtain two distinct clusters. The silhouette coefficient is computed as an internal evaluation criterion for the quality of the clustering, and has an average value of 0.81. In order to confirm that the two clusters correspond to QS and non-QS periods, external evaluation of the clustering using the visual sleep stage scoring by 2 clinicians is performed as well. Comparison with the clinical labels revealed that the neural complexity is significantly lower in QS compared to non-QS. This prior information is used to label the clusters and the average sensitivity, specificity, accuracy and area under the curve are equal to 72%, 80%, 78% and 84%, respectively. This study shows that the complexity of brain dynamics exhibits fundamental differences between vigilance states in preterm newborns. Moreover, a novel unsupervised approach to detect quiet sleep based on the tensor factorization of the multiscale entropy tensor is proposed.
15 mins
Brain-Heart Interaction to Assess Stress in Premature Infants
Mario Lavanga, Chiara Bernagie, Alexander Caicedo, Katrien Jansen, Els Ortibus, Gunnar Naulaers, Sabine Van Huffel
Abstract: Due to the early stress in the neonatal intensive care unit, premature patients might be prone to delayed brain maturation. Some example of stressor events are apneas or long bradycardias, which disrupt the sleep-wake cycle [1]. The goal of this research is to detect stress in preterm infants based on EEG activity, heart-rate variability (HRV) and their interaction. Multichannel EEG and ECG were recorded from 29 premature patients during 4 hours, while the stress levels were assessed using the Leuven Pain Score. After the HRV computation, bradycardia events were located and the associated EEG δ oscillations were extracted via an analytic wavelet transform, using a window starting 3 min before the bradycardia until three min afterwards. We used standard time and spectral features, in addition, the coupling between the δ oscillations and the HRV (EEG-HRV) and the connectivity among EEG channels (EEG-EEG) were computed via analytic Wavelet Coherence and the time-delay stability analysis [2]. A least-squares support vector machines (LS-SVM) classifier was implemented to discriminate between bradycardias during stress and in its absence. The main conclusion is that brain-heart (EEG-HRV) coupling outperforms other methodologies in the classification of the bradycardias. This is shown by AUC metrics for the different features: 0.76 for EEG-HRV, 0.71 for EEG-EEG connectivity, 0.69 for univariate EEG, 0.63 for univariate HRV. Furthermore, an increase in connectivity during the bradycardia event and its enhancement with higher stress levels indicates an impact of the autonomic system on the brain, which could affect the patient’s normal development.
15 mins
Monitoring Infant Brain Perfusion by Trans-Fontanel Echography (MIFFY)
Jorinde Kortenbout, Rik Vos, Jeroen Dudink, Paul Govaert, Martin Verweij, Michiel Pertijs, Nico de Jong
Abstract: Very preterm neonates (24-32 weeks of gestation) are born in a critical period of brain development and maturation. The brain is extremely vulnerable in this period and injuries during this phase may lead to long term cognitive, motor and behavioural problems. Adequate brain perfusion is important in prevention of preterm brain injury [1]. Objective continuous monitoring of brain perfusion is not yet possible in the neonatal intensive care (NICU). In the NICU, conventional ultrasound is used for evaluation of the neonatal brain anatomy and detection of brain injury [2]. High frame rate (HFR) ultrasound (>1000 Hz) enables sensitive vascular imaging and non-invasive elasticity imaging [3]. To permit continuous HFR neonatal brain monitoring a trans-fontanel ultrasound probe (MIFFY probe) will be developed. Volumetric (3D) ultrasound data of the premature brain will be acquired in HFR to obtain high resolution grey scale images, perfusion images of the vascular tree and elasticity images. RF data will be collected with a wide opening angle and a broad frequency band through the anterior fontanel every ten minutes, during at least one cardiac cycle. The high frequency probe makes it possible to image and quantify flow in vessels with a size of 100-200 µm. The method is sensitive for slow flow as well and even subtle changes in hemodynamics can be detected. Since the data will be acquired every ten minutes, the infant can act as its own reference, so that an alarm can be given if the perfusion changes significantly. The dataset will also be used to generate elasticity images. These images give information about local stiffness of the preterm brain. We are currently developing this technique in stiffness phantoms. The aim is to explore the natural occurring shear waves (heartbeat, breathing) and this might provide information of changes in tissue stiffness after perfusion deficiencies. The MIFFY probe with high resolution will have the ability to improve diagnostic value, since minor changes in perfusion or elasticity can be measured in a timely manner to provide continuous neuromonitoring in the most critical days of neonatal intensive care.
15 mins
Improving Maternal Care in Resource-Limited Settings Using a Low-Cost Ultrasound Device and Machine Learning
Thomas van den Heuvel, Bram van Ginneken, Chris de Korte
Abstract: Worldwide, 99% of all maternal deaths occur in developing countries. Ultrasound is normally used to detect maternal risk factors, but it is rarely available in developing countries because it is too expensive, and it requires a trained sonographer to acquire and interpret the ultrasound images. We use a low-cost ultrasound device which was combined with the obstetric sweep protocol (OSP) and deep learning algorithms to automatically detect maternal risk factors. The OSP can be taught to any health care worker without prior knowledge of ultrasound within one day. The OSP was acquired from 318 pregnant women using the low-cost MicrUs (Telemed Ultrasound Medical Systems, Milan, Italy) in Ethiopia. Two deep learning networks and two random forest classifiers were trained to automatically detect twin pregnancies, estimate gestational age (GA) and determine fetal presentation. The first deep learning network performs a frame classification, which was used to automatically separate the six sweeps of the OSP and automatically detect the presence of the fetal head and torso in each frame. The second deep learning network was trained to measure the head circumference (HC) using all frames in which the first deep learning system detected the presence of a fetal head. The HC was used to determine the GA. Two random forest classifiers were trained to detect twin pregnancies and determine fetal presentation using the frame classification of the first deep learning network. The developed algorithm can automatically estimate the GA with an interquartile range of 15.2 days, correctly detected 61% of all twins with a specificity of 99%, and correctly detect all 31 breech presentations and 215 of the 216 cephalic presentations. The developed algorithm can be computed in less than two seconds, making real-time analysis of the OSP feasible. The presented system is able to determine three maternal risk factors using the OSP. The OSP can be acquired without the need of a trained sonographer, which makes widespread obstetric ultrasound affordable and fast to implement in resource-limited settings. This makes is possible to refer pregnant women in time to a hospital to receive treatment when risk factors are detected.
15 mins
Non-Invasive Multi-Parametric Ultrasound (NIMPUS) Classification Tool to Assess Fetal Lung Maturity
Gert Weijers, Freke Wilmink, Frank van der Bussche, Chris de Korte
Abstract: Neonatal Respiratory Morbidity (NRM) is highly associated with fetal lung in-maturity and associated with late premature (34+0-36+6) or early term (37+0-38+6) delivery of the newborn. The risk at NRM and the maternal capacity and risks must be balanced against each other in the decision making concerning an elective delivery. Currently, lung maturity is hard to assess, and not feasible using non-invasive tools. Therefore a noninvasive diagnostic tool is highly warrant to support this decision making. Since ultrasound is non-invasive and already widely used in gynaecology, three Voluson ultrasound machines (GE Healthcare, 1x E8 and 2x E10, using RM6C transducer bandwidth: 1-7MHz) were used to image the fetal lung, using fixed and calibrated imaging preset. The calibrations were performed using a Tissue Mimicking Phantom (TMP) and included: optimal zoom-setting for fully speckle sampling; beam-profile correction estimation; scaling of all texture parameters relative to the TMP used, and to scale US parameters from one to another machine. Three independent conventional ultrasound images of the fetal thorax (four chamber view) of 45 babies from singleton pregnant women were acquired (gestational age range: 20-39 weeks). The CAUS software (Weijers, Wanten et al. 2016) was used to estimate the mean echo level, residual attenuation coefficient (RAC), and 2D parametric texture images of the axial (AX) and lateral (LAT) speckle size [mm] and amplitudes (_max) [dB]. All parameters were calculated after manual annotation of the lung parenchyma, and correlated to gestation age in order to search for US parameters able to predict the gestational and thus stage the long-ripeness. Highly significant correlations for all texture entropy parameters (AX; AX_max; LAT; LAT_max: R = .56**; .66**; .62**; .51** respectively) with gestational age were found. Also the RAC (R=-.38**) and the lung area (R=0.75**) correlated well with gestational age. These results indicate the potential of NIMPUS for fetal lung ripeness staging. Further research including, addition of NRM scores of the newborns and performing multiparametric logistics regression analysis, have to be conducted to obtain cut-off values required for predictive values estimation for NIMPUS. REFERENCE: Weijers, G., G. Wanten, J. M. Thijssen, M. van der Graaf and C. L. de Korte (2016). "Quantitative Ultrasound for Staging of Hepatic Steatosis in Patients on Home Parenteral Nutrition Validated with Magnetic Resonance Spectroscopy: A Feasibility Study." Ultrasound Med Biol 42(3): 637-644.
15 mins
Nonlinear Transfer Entropy to Assess the Neurometabolic Coupling in Premature Neonates
Dries Hendrikx, Liesbeth Thewissen, Anne Smits, Gunnar Naulaers, Karel Allegaert, Sabine Van Huffel, Alexander Caicedo
Abstract: Neurometabolic coupling (NMC) is a regulation mechanism that helps the brain to maintain an appropriate energy flow to the neural tissue under conditions of increased neuronal activity. It is however not yet clear to which extent this mechanism is present in the premature brain. In this study, we explore the use of transfer entropy (TE) in order to compute the nonlinear coupling between changes in brain function, assessed by means of EEG, and changes in brain oxygenation, assessed by means of near-infrared spectroscopy (NIRS). In a previous study, we have measured the coupling between both variables using a linear model to compute TE. The results indicated that changes in brain oxygenation were likely to precede changes in EEG activity. However, using a nonlinear and nonparametric approach to compute TE, the results indicate an opposite directionality of this coupling. The source of the different results provided by the linear and nonlinear TE is unclear and needs further research. We present the results from a cohort of 21 premature neonates, which are stratified in two groups, based on the presence of brain abnormalities observed on routine ultrasound (US) scans. Results indicate that TE values computed using the nonlinear approach are able to discriminate neonates with brain abnormalities from healthy neonates, indicating a possibly less functional NMC in neonates with brain abnormalities.