Manon J.N.L. Benders

Automatic Segmentation of MR Brain Images With a Convolutional Neural Network

Automatic segmentation in MR brain images is important for quantitative analysis in large-scale studies with images acquired at all ages. This paper presents a method for the automatic segmentation of MR brain images into a number of tissue classes using a convolutional neural network. To ensure that the method obtains accurate segmentation details as well as spatial consistency, the network uses multiple patch sizes and multiple convolution kernel sizes to acquire multi-scale information about each voxel. The method is not dependent on explicit features, but learns to recognise the information that is important for the classification based on training data. The method requires a single anatomical MR image only. The segmentation method is applied to five different data sets: coronal T2-weighted images of preterm infants acquired at 30 weeks postmenstrual age (PMA) and 40 weeks PMA, axial T2-weighted images of preterm infants acquired at 40 weeks PMA, axial T1-weighted images of ageing adults acquired at an average age of 70 years, and T1-weighted images of young adults acquired at an average age of 23 years. The method obtained the following average Dice coefficients over all segmented tissue classes for each data set, respectively: 0.87, 0.82, 0.84, 0.86, and 0.91. The results demonstrate that the method obtains accurate segmentations in all five sets, and hence demonstrates its robustness to differences in age and acquisition protocol.Automatic segmentation in MR brain images is important for quantitative analysis in large-scale studies with images acquired at all ages. This paper presents a method for the automatic segmentation of MR brain images into a number of tissue classes using a convolutional neural network. To ensure that the method obtains accurate segmentation details as well as spatial consistency, the network uses multiple patch sizes and multiple convolution kernel sizes to acquire multi-scale information about each voxel. The method is not dependent on explicit features, but learns to recognise the information that is important for the classification based on training data. The method requires a single anatomical MR image only. The segmentation method is applied to five different data sets: coronal T2-weighted images of preterm infants acquired at 30 weeks postmenstrual age (PMA) and 40 weeks PMA, axial T2-weighted images of preterm infants acquired at 40 weeks PMA, axial T1-weighted images of ageing adults acquired at an average age of 70 years, and T1-weighted images of young adults acquired at an average age of 23 years. The method obtained the following average Dice coefficients over all segmented tissue classes for each data set, respectively: 0.87, 0.82, 0.84, 0.86, and 0.91. The results demonstrate that the method obtains accurate segmentations in all five sets, and hence demonstrates its robustness to differences in age and acquisition protocol.

The Neonatal Connectome During Preterm Brain Development

The human connectome is the result of an elaborate developmental trajectory. Acquiring diffusion-weighted imaging and resting-state fMRI, we studied connectome formation during the preterm phase of macroscopic connectome genesis. In total, 27 neonates were scanned at week 30 and/or week 40 gestational age (GA). Examining the architecture of the neonatal anatomical brain network revealed a clear presence of a small-world modular organization before term birth. Analysis of neonatal functional connectivity (FC) showed the early formation of resting-state networks, suggesting that functional networks are present in the preterm brain, albeit being in an immature state. Moreover, structural and FC patterns of the neonatal brain network showed strong overlap with connectome architecture of the adult brain (85 and 81%, respectively). Analysis of brain development between week 30 and week 40 GA revealed clear developmental effects in neonatal connectome architecture, including a significant increase in white matter microstructure (P < 0.01), small-world topology (P < 0.01) and interhemispheric FC (P < 0.01). Computational analysis further showed that developmental changes involved an increase in integration capacity of the connectivity network as a whole. Taken together, we conclude that hallmark organizational structures of the human connectome are present before term birth and subject to early development.

Myth: Cerebral palsy cannot be predicted by neonatal brain imaging

There is controversy in the literature about the value of brain imaging in neonates regarding the prediction of cerebral palsy (CP). The aim of this review was to unravel the myth that CP cannot be predicted by neuroimaging in neonates. Major intracranial lesions in the preterm infant should be recognized with sequential cranial ultrasound and will predict those with non-ambulatory CP. Magnetic resonance imaging (MRI) at term-equivalent age will refine the prediction by assessment of myelination of the posterior limb of the internal capsule. Prediction of motor outcome in preterm infants with subtle white matter injury remains difficult, even with conventional MRI. MRI is a better tool to predict outcome in the term infant with hypoxic-ischaemic encephalopathy or neonatal stroke. The use of diffusion-weighted imaging as an additional sequence adds to the predictive value for motor outcome. Sequential and dedicated neuroimaging should enable us to predict motor outcome in high risk newborns infants.

Cerebral near infrared spectroscopy oximetry in extremely preterm infants : Phase II randomised clinical trial

Objective To determine if it is possible to stabilise the cerebral oxygenation of extremely preterm infants monitored by cerebral near infrared spectroscopy (NIRS) oximetry. Design Phase II randomised, single blinded, parallel clinical trial. Setting Eight tertiary neonatal intensive care units in eight European countries. Participants 166 extremely preterm infants born before 28 weeks of gestation: 86 were randomised to cerebral NIRS monitoring and 80 to blinded NIRS monitoring. The only exclusion criterion was a decision not to provide life support. Interventions Monitoring of cerebral oxygenation using NIRS in combination with a dedicated treatment guideline during the first 72 hours of life (experimental) compared with blinded NIRS oxygenation monitoring with standard care (control). Main outcome measures The primary outcome measure was the time spent outside the target range of 55-85% for cerebral oxygenation multiplied by the mean absolute deviation, expressed in %hours (burden of hypoxia and hyperoxia). One hour with an oxygenation of 50% gives 5%hours of hypoxia. Secondary outcomes were all cause mortality at term equivalent age and a brain injury score assessed by cerebral ultrasonography. Randomisation Allocation sequence 1:1 with block sizes 4 and 6 in random order concealed for the investigators. The allocation was stratified for gestational age (<26 weeks or ≥26 weeks). Blinding Cerebral oxygenation measurements were blinded in the control group. All outcome assessors were blinded to group allocation. Results The 86 infants randomised to the NIRS group had a median burden of hypoxia and hyperoxia of 36.1%hours (interquartile range 9.2-79.5%hours) compared with 81.3 (38.5-181.3) %hours in the control group, a reduction of 58% (95% confidence interval 35% to 73%, P<0.001). In the experimental group the median burden of hypoxia was 16.6 (interquartile range 5.4-68.1) %hours, compared with 53.6 (17.4-171.3) %hours in the control group (P=0.0012). The median burden of hyperoxia was similar between the groups: 1.2 (interquartile range 0.3-9.6) %hours in the experimental group compared with 1.1 (0.1-23.4) %hours in the control group (P=0.98). We found no statistically significant differences between the two groups at term corrected age. No severe adverse reactions were associated with the device. Conclusions Cerebral oxygenation was stabilised in extremely preterm infants using a dedicated treatment guideline in combination with cerebral NIRS monitoring. Trial registration ClinicalTrial.gov NCT01590316.

Maternal and Infant Characteristics Associated With Perinatal Arterial Stroke in the Preterm Infant

Background and Purpose— Most perinatal arterial stroke (PAS) studies that investigated infant characteristics have excluded preterm infants from the study population. We sought to analyze the imaging findings and antenatal and perinatal risk factors in preterm infants with PAS. Methods— This was a hospital-based, case-control study of preterm infants. Case infants were confirmed by reviewing brain imaging scans and medical records (n=31). Three controls per case were individually matched with case infants from the study population. Results— Gestational age ranged between 27 and 36 weeks, and birth weight ranged between 580 and 3180 g. PAS was more common on the left side (61%), and 7% had bilateral PAS. The majority of strokes involved the middle cerebral artery distribution. Involvement of 1 or more lenticulostriate branches was most common among infants with a gestational age of 28 to 32 weeks, but main branch involvement was seen only in those with a gestational age of >32 weeks. Twin-to-twin-transfusion syndrome, fetal heart rate abnormality, and hypoglycemia were identified as independent risk factors for PAS. Conclusions— Preterm PAS is associated with prenatal, perinatal, and postpartum risk factors. We were unable to identify any maternal risk factors. Involvement of the different branches of the middle cerebral artery changed with an increase in gestational age.

Structural asymmetries of perisylvian regions in the preterm newborn

During the last trimester of human pregnancy, the cerebral cortex of foetuses becomes greatly and quickly gyrified, and post-mortem studies have demonstrated that hemispheres are already asymmetric at the level of Heschl gyrus, planum temporale and superior temporal sulcus (STS). Recently, magnetic resonance imaging (MRI) and dedicated post-processing tools enabled the quantitative study of brain development non-invasively in the preterm newborn. However, previous investigations were conducted either over the whole brain or in specific sulci. These approaches may consequently fail to highlight most cerebral sites, where anatomical landmarks are hard to delineate among individuals. In this cross-sectional study, we aimed to blindly and automatically map early asymmetries over the immature cortex. Voxel-based analyses of cortical and white matter masks were performed over a group of 25 newborns from 26 to 36 weeks of gestational age. Inter-individual variations associated with increasing age were first detected in large cerebral regions, with a prevalence of the right hemisphere in comparison with the left. Asymmetries were further highlighted in three specific cortical regions. Confirming previous studies, we observed deeper STS on the right side and larger posterior region of the sylvian fissure on the left side, close to planum temporale. For the first time, we also detected larger anterior region of the sylvian fissure on the left side, close to Brocas region. This study demonstrated that perisylvian regions are the only regions to be asymmetric from early on, suggesting their anatomical specificity for the emergence of functional lateralization in language processing prior to language exposure.

Cerebellar Volume and Proton Magnetic Resonance Spectroscopy at Term, and Neurodevelopment at 2 Years of Age in Preterm Infants.

Aim To assess the relation between cerebellar volume and spectroscopy at term equivalent age, and neurodevelopment at 24 months corrected age in preterm infants.

MR Imaging and Outcome of Term Neonates with Perinatal Asphyxia: Value of Diffusion-weighted MR Imaging and H MR Spectroscopy

PURPOSE To compare the association between neurodevelopmental outcome in neonates with hypoxic-ischemic encephalopathy following perinatal asphyxia and (a) apparent diffusion coefficients (ADCs) in the thalamus and basal ganglia at diffusion-weighted (DW) magnetic resonance (MR) imaging and (b) hydrogen 1 (¹H) MR spectroscopic measurements in the basal ganglia. MATERIALS AND METHODS This retrospective study was approved by the local ethics committee, and the requirement to obtain informed consent was waived. Eighty-one term neonates with perinatal asphyxia underwent conventional and DW cranial MR imaging (median age, 4 days; age range, 1-14 days); 51 neonates also underwent ¹H MR spectroscopy. Neurodevelopment was assessed from 18 to 46 months. Patients with favorable and adverse outcomes were compared. Receiver operating characteristics analysis was performed in all patients, and uni- and multivariate logistic regression analyses were performed in 44 patients examined within 7 days of birth by using MR imaging scores, ADCs in the basal ganglia and thalamus, and ¹H MR spectroscopic measurements in the basal ganglia. RESULTS An adverse outcome was seen in 28 of all 81 neonates (20 died, seven developed cerebral palsy, and one had severe mental retardation) and 22 of the 44 neonates examined within 7 days of birth with both ADC and ¹H MR spectroscopy. Poor outcome was associated with (a) lower ADCs in the basal ganglia (P < .001) and thalamus (P = .001) of neonates examined within 7 days of birth and (b) a higher lactate (Lac)-N-acetylaspartate (NAA) ratio in the basal ganglia (P < .001). Multivariate analysis showed that MR imaging score combined with Lac/NAA ratios or ADCs in the basal ganglia within the 1st week of life had a better association with outcome than did MR imaging alone (P = .006, area under the receiver operating characteristic curve [AUC] = 0.85 with Lac/NAA ratio; P < .0001, AUC = 0.93 with ADCs in basal ganglia). CONCLUSION The combination of MR imaging score with ADCs or Lac/NAA ratios in the basal ganglia has a better association with outcome of asphyxiated term neonates than does MR imaging alone.

Impact of nutrition on brain development and its neuroprotective implications following preterm birth

The impact of nutrition on brain development in preterm infants has been increasingly appreciated. Early postnatal growth and nutrient intake have been demonstrated to influence brain growth and maturation with subsequent effects on neurodevelopment that persist into childhood and adolescence. Nutrition could also potentially protect against injury. Inflammation and perinatal infection play a crucial role in the pathogenesis of white matter injury, the most common pattern of brain injury in preterm infants. Therefore, nutritional components with immunomodulatory and/or anti-inflammatory effects may serve as neuroprotective agents. Moreover, growing evidence supports the existence of a microbiome-gut-brain axis. The microbiome is thought to interact with the brain through immunological, endocrine, and neural pathways. Consequently, nutritional components that may influence gut microbiota may also exert beneficial effects on the developing brain. Based on these properties, probiotics, prebiotic oligosaccharides, and certain amino acids are potential candidates for neuroprotection. In addition, the amino acid glutamine has been associated with a decrease in infectious morbidity in preterm infants. In conclusion, early postnatal nutrition is of major importance for brain growth and maturation. Additionally, certain nutritional components might play a neuroprotective role against white matter injury, through modulation of inflammation and infection, and may influence the microbiome-gut-brain axis.

Risk factors for perinatal arterial ischaemic stroke in full-term infants: a case-control study

Objective The incidence of perinatal arterial ischaemic stroke (PAIS) is about 1 in 2300 live births. Evidence about the aetiology is still lacking. The aim of this study was to identify maternal, perinatal and neonatal risk factors for symptomatic PAIS in full-term infants. Methods Each full-term infant with PAIS was matched to three healthy controls for gestational age, date of birth and hospital of birth. Antenatal and perinatal risk factors were studied using univariate and multivariate conditional logistic regression analysis. Results Fifty-two infants were diagnosed with PAIS. Significant risk factors in the univariate analysis (p<0.05) were nulliparity (64% vs 47%), maternal fever (>38°C) during delivery (10% vs 1%), fetal heart rate decelerations (63% vs 16%), meconium-stained amniotic fluid (44% vs 17%), emergency caesarean section (35 vs 2%), Apgar score (1 min) ≤3 (29% vs 1%), Apgar score (5 min) <7 (25% vs 1%), umbilical artery pH <7.10 (56% vs 10%), hypoglycaemia <2.0 mmol/l (29% vs 3%) and early-onset sepsis/meningitis (14% vs 2%). In the multivariate analysis, maternal fever (OR 10.2; 95% CI 1.3 to 78.5), Apgar score (5 min) <7 (OR 18.1; 95% CI 3.4 to 96.8), hypoglycaemia <2.0 mmol/l (OR 13.0; 95% CI 3.2 to 52.6) and early-onset sepsis/meningitis (OR 5.8; 95% CI 1.1 to 31.9) were significantly associated with PAIS. Conclusions Maternal fever during delivery and early-onset sepsis/meningitis were found to be involved with PAIS as was previously noted. Apgar score (5 min) <7 and hypoglycaemia were found to be important risk factors in term PAIS.