Automated identification of abnormal infant movements from smart phone videos.

Cerebral palsy (CP) is the most common cause of physical disability during childhood, occurring at a rate of 2.1 per 1000 live births. Early diagnosis is key to improving functional outcomes for children with CP. The General Movements (GMs) Assessment has high predictive validity for the detection of CP and is routinely used in high-risk infants but only 50% of infants with CP have overt risk factors when they are born. The implementation of CP screening programs represents an important endeavour, but feasibility is limited by access to trained GMs assessors. To facilitate progress towards this goal, we report a deep-learning framework for automating the GMs Assessment. We acquired 503 videos captured by parents and caregivers at home of infants aged between 12- and 18-weeks term-corrected age using a dedicated smartphone app. Using a deep learning algorithm, we automatically labelled and tracked 18 key body points in each video. We designed a custom pipeline to adjust for camera movement and infant size and trained a second machine learning algorithm to predict GMs classification from body point movement. Our automated body point labelling approach achieved human-level accuracy (mean ± SD error of 3.7 ± 5.2% of infant length) compared to gold-standard human annotation. Using body point tracking data, our prediction model achieved a cross-validated area under the curve (mean ± S.D.) of 0.80 ± 0.08 in unseen test data for predicting expert GMs classification with a sensitivity of 76% ± 15% for abnormal GMs and a negative predictive value of 94% ± 3%. This work highlights the potential for automated GMs screening programs to detect abnormal movements in infants as early as three months term-corrected age using digital technologies.

Postnatal Brain Growth Assessed by Sequential Cranial Ultrasonography in Infants Born <30 Weeks' Gestational Age.

BACKGROUND AND PURPOSE: Brain growth in the early postnatal period following preterm birth has not been well described. This study of infants born at <30 weeks' gestational age and without major brain injury aimed to accomplish the following: 1) assess the reproducibility of linear measures made from cranial ultrasonography, 2) evaluate brain growth using sequential cranial ultrasonography linear measures from birth to term-equivalent age, and 3) explore perinatal predictors of postnatal brain growth. MATERIALS AND METHODS: Participants comprised 144 infants born at <30 weeks' gestational age at a single center between January 2011 and December 2013. Infants with major brain injury seen on cranial ultrasonography or congenital or chromosomal abnormalities were excluded. Brain tissue and fluid spaces were measured from cranial ultrasonography performed as part of routine clinical care. Brain growth was assessed in 3 time intervals: <7, 7-27, and >27 days' postnatal age. Data were analyzed using intraclass correlation coefficients and mixed-effects regression. RESULTS: A total of 429 scans were assessed for 144 infants. Several linear measures showed excellent reproducibility. All measures of brain tissue increased with postnatal age, except for the biparietal diameter, which decreased within the first postnatal week and increased thereafter. Gestational age of ≥28 weeks at birth was associated with slower growth of the biparietal diameter and ventricular width compared with gestational age of <28 weeks. Postnatal corticosteroid administration was associated with slower growth of the corpus callosum length, transcerebellar diameter, and vermis height. Sepsis and necrotizing enterocolitis were associated with slower growth of the transcerebellar diameter. CONCLUSIONS: Postnatal brain growth in infants born at <30 weeks' gestational age can be evaluated using sequential linear measures made from routine cranial ultrasonography and is associated with perinatal predictors of long-term development.

Early MRI in term infants with perinatal hypoxic-ischaemic brain injury: interobserver agreement and MRI predictors of outcome at 2 years.

AIM: To compare diffusion-weighted imaging (DWI) and non-DWI magnetic resonance imaging (MRI), proton MR spectroscopy (1H-MRS), and clinical biomarkers for prediction of 2 year developmental outcome in term infants with perinatal hypoxic-ischaemic encephalopathy (HIE). MATERIALS AND METHODS: Nineteen infants ≥36 weeks gestation with HIE were recruited and MRI performed day 3-7 (mean = 5). MRI was scored independently by three radiologists using a standardized scoring system. Lactate-to-N-acetylaspartate ratio (Lac:NAA) in the lentiform nucleus was calculated. Developmental assessment was performed at 2 years using the Bayley Scales of Infant and Toddler Development (BSID-III). Interobserver agreement about abnormality in 10 brain regions was measured. Univariate analysis was performed to determine variables associated with adverse outcome (i.e., death or Bayley score for any domain <70). RESULTS: Good interobserver agreement (kappa = 0.61-0.69) on scores for DWI was obtained for the cortex, putamen, and brainstem, but not for any region on non-DWI. A significant association was found between outcome and Lac:NAA (p < 0.003) and DWI scores for lentiform nucleus, thalamus, cortex, posterior limb of the internal capsule (PLIC), and paracentral white matter (p = 0.001-0.013), but for non-DWI score only in the vermis or brainstem. A combination of Lac:NAA ≥0.25 or DWI/apparent diffusion coefficient (ADC) signal abnormality in the PLIC had 100% specificity and sensitivity for poor outcome. CONCLUSION: Interobserver agreement for non-DWI performed during the first week is poor. Agreement by three radiologists about the presence of abnormal signal within the PLIC on ADC/DWI images or elevation of Lac:NAA above 0.25 improved sensitivity without reducing the prognostic specificity of MRS in the 19 patients, but this requires validation in a larger group of infants with HIE who have been treated with hypothermia.

Extremely preterm birth and adolescent mental health in a geographical cohort born in the 1990s.

BACKGROUND: Preterm birth confers risk for poor outcome, including mental health problems. Survival of extremely preterm (EP; <28 weeks' gestation) or extremely low birthweight (ELBW; <1000 g) infants increased in the 1990s but psychiatric outcomes in older adolescents born preterm since 1990 are not well documented. This study aimed to characterize mental health and personality traits in a prospective geographical cohort of adolescents born EP/ELBW in Victoria, Australia in 1991 and 1992. METHOD: At age 18 years, 215 EP/ELBW and 157 normal birthweight (>2499 g) control adolescents completed the Structured Clinical Interview for DSM-IV Disorders, Axis 1 Non-Patient version (SCID-I/NP), the Children's Interview for Psychiatric Syndromes (ChIPS) attention deficit hyperactivity disorder (ADHD) module, and questionnaires assessing recent depression and anxiety symptoms and personality traits. RESULTS: ADHD prevalence was significantly elevated in EP/ELBW adolescents compared with controls [15% v. 7%; odds ratio (OR) 2.67, 95% confidence interval (CI) 1.08-6.58]. Aside from ADHD, however, EP/ELBW and control adolescents reported very similar outcomes, with other lifetime diagnoses identified in 23% of EP/ELBW and 21% of controls. These were predominantly mood and anxiety disorders (21% EP/ELBW, 20% controls). The groups did not differ in recent depression or anxiety symptoms assessed using questionnaires, and personality traits were also similar. CONCLUSIONS: ADHD was more prevalent in EP/ELBW adolescents than controls, which is consistent with some, but not all, reports on preterm survivors born before the 1990s, and younger preterm children born in the 1990s. The high rates of anxiety and mood disorders were similar in both groups, and comparable with population-based estimates.

Postnatal ultrasound reliability in cerebellar vermis assessment.

BACKGROUND: Cerebellar growth in late gestation is impeded by prematurity which may adversely affect neurocognitive development. Assessment of cerebellar growth should be easily attainable, reliable and reproducible. OBJECTIVE: To assess the reliability of linear sonographic cerebellar vermis measurement. METHOD: Cranial ultrasounds of 110 infants ranging from 24 to 41 weeks' gestation were retrospectively reviewed. Cerebellar vermian height, craniocaudal diameter and superior and inferior vermis widths were independently measured on the first midline sagittal image by three neonatal sonologists of varying experience. Interobserver and intraobserver reliability were calculated using the intraclass correlation coefficient (ICC) (2 way mixed model, SPSS V.15.0). RESULTS: 61 images were technically adequate. Interobserver ICCs (95% CI) were: cerebellar vermian height 0.88 (0.82 to 0.92); craniocaudal diameter 0.91 (0.86 to 0.94); superior vermis width 0.84 (0.77 to 0.89); inferior vermis width 0.92 (0.89 to 0.95). Intraobserver ICCs were similar. CONCLUSION: With adequate images, linear ultrasound measurements of cerebellar vermis are reliable.

Abnormal white matter signal on MR imaging is related to abnormal tissue microstructure.

BACKGROUND AND PURPOSE: White matter signal-intensity abnormalities (WMSA) on MR imaging are related to adverse neurodevelopmental outcome in extremely preterm infants. Diffusion tensor imaging (DTI) may detect alterations in cerebral white matter microstructure and thus may help confirm the pathologic basis of WMSA. This study aimed to relate regional DTI measures with severity of WMSA in very preterm infants. MATERIALS AND METHODS: One hundred eleven preterm infants (birth weight, <1250 g and/or gestational age, <30 weeks) were scanned at term-equivalent age (1.5T). WMSA were classified as normal, focal, or extensive. Apparent diffusion coefficient (ADC), fractional anisotropy (FA), axial (lambda1), and radial ([lambda2 + lambda3]/2) diffusivity were calculated in 12 regions of interest placed in the bilateral posterior limbs of the internal capsule, frontal (superior and inferior), sensorimotor, and occipital (superior and inferior) white matter regions. Data were compared by using 1-way analysis of variance, with a Bonferroni correction for multiple comparisons. RESULTS: Thirty-nine infants had normal, 59 infants had focal, and 13 infants had extensive WMSA. Compared with infants with normal or focal WMSA, infants with extensive WMSA had significantly lower FA in the internal capsule (P < .001), right inferior frontal regions (P < .05), and right superior occipital regions (P = .01); and higher radial diffusivity in the right internal capsule (P = .005), bilateral sensorimotor (P < .05), and right superior occipital regions (P < .05). Compared with infants with normal WMSA, infants with extensive WMSA had significantly higher ADC in bilateral sensorimotor regions (P < .01) and right superior occipital regions (P = .01), and lower axial diffusivity in the bilateral sensorimotor regions (P < .05). CONCLUSIONS: There are significant region-specific changes in ADC, FA, radial diffusivity, and axial diffusivity in preterm infants with extensive WMSA. Altered radial diffusivity was most prominent. This implies that disrupted premyelinating oligodendroglia is the major correlate with extensive WMSA rather than axonal pathology.

Proton MR spectroscopy in neonates with perinatal cerebral hypoxic-ischemic injury: metabolite peak-area ratios, relaxation times, and absolute concentrations.

BACKGROUND: Results from cerebral proton (1)H-MR spectroscopy studies of neonates with perinatal hypoxic-ischemic injury have generally been presented as metabolite peak-area ratios, which are T1- and T2-weighted, rather than absolute metabolite concentrations. We hypothesized that compared with (1)H-MR spectroscopy peak-area ratios, calculation of absolute metabolite concentrations and relaxation times measured within the first 4 days after birth (1) would improve prognostic accuracy and (2) enhance the understanding of underlying neurochemical changes in neonates with neonatal encephalopathy. METHODS: Seventeen term infants with neonatal encephalopathy and 10 healthy controls were studied at 2.4T at 1 (1-3) and 2 (2-4) (median [interquartile range]) days after birth, respectively. Infants with neonatal encephalopathy were classified into 2 outcome groups (normal/mild and severe/fatal), according to neurodevelopmental assessments at 1 year. The MR spectroscopy peak-area ratios, relaxation times, absolute concentrations, and concentration ratios of lactate (Lac), creatine plus phosphocreatine (Cr), N-acetylaspartate (NAA), and choline-containing compounds (Cho) from a voxel centered on the thalami were analyzed according to outcome group. RESULTS: Comparing the severe/fatal group with the controls (significance assumed with P < 0.05), we found that Lac/NAA, Lac/Cho, and Lac/Cr peak-area ratios increased and NAA/Cr and NAA/Cho decreased; Lac, NAA, and Cr T2s were increased; [Lac] was increased and [Cho], [Cr], and [NAA] decreased; and among the concentration ratios, only [Lac]/[NAA] was increased. Comparison of the normal/mild group with controls revealed no differences in peak-area ratios, relaxation times, or concentration ratios but decreased [NAA], [Cho], and [Cr] were observed in the infants with normal/mild outcome. Comparison of the normal/mild and severe/fatal groups showed increased Lac/NAA and Lac/Cho and decreased NAA/Cr and NAA/Cho peak-area ratios, reduced [NAA], and increased Lac T2 in the infants with the worse outcome. CONCLUSIONS: Metabolite concentrations, in particular [NAA], enhance the prognostic accuracy of cerebral (1)H-MR spectroscopy-[NAA] was the only measurable to discriminate among all (control, normal/mild, and severe/fatal outcome) groups. However, peak-area ratios are more useful prognostic indicators than concentration ratios because they depend on metabolite concentrations and T2s, both of which are pathologically modulated. Concentration ratios depend only on the concentrations of the constituent metabolites. Increased Cr T2 may provide an indirect marker of impaired cellular energetics, and similarly, NAA T2 may constitute an index of exclusively neuronal energy status. Our recommendation is to collect data that enable calculation of brain metabolite concentrations. However, if time constraints make this impossible, metabolite peak-area ratios provide the next best method of assigning early prognosis in neonatal encephalopathy.

Gastrointestinal manifestations of postnatal cytomegalovirus infection in infants admitted to a neonatal intensive care unit over a five year period.

Sixteen cases of postnatal cytomegalovirus (CMV) infection were identified in a neonatal intensive care unit population over a five year period. Eleven of these infants had gastrointestinal signs at the time of presentation. These ranged from minor and transient (abdominal distension and enteral feed intolerance) to severe and life threatening (protein losing enteropathy, diarrhoea, and hypernatraemic dehydration). An initial diagnosis of necrotising enterocolitis was common, but no infant showed intestinal or hepatic portal pneumatosis. The gestational age of the infants was 24-38 weeks. All had received fresh maternal breast milk. It is suggested that CMV enteritis is added to the spectrum of clinical manifestations of postnatal CMV infection. Signs suggestive of necrotising enterocolitis with atypical features should prompt investigations for CMV infection.