Deep Learning to Optimize Magnetic Resonance Imaging Prediction of Motor Outcomes After Hypoxic-Ischemic Encephalopathy.

BACKGROUND: Magnetic resonance imaging (MRI) is the gold standard for outcome prediction after hypoxic-ischemic encephalopathy (HIE). Published scoring systems contain duplicative or conflicting elements. METHODS: Infants ≥36 weeks gestational age (GA) with moderate to severe HIE, therapeutic hypothermia treatment, and T1/T2/diffusion-weighted imaging were identified. Adverse motor outcome was defined as Bayley-III motor score <85 or Alberta Infant Motor Scale <10th centile at 12 to 24 months. MRIs were scored using a published scoring system. Logistic regression (LR) and gradient-boosted deep learning (DL) models quantified the importance of clinical and imaging features. The cohort underwent 80/20 train/test split with fivefold cross validation. Feature selection eliminated low-value features. RESULTS: A total of 117 infants were identified with mean GA = 38.6 weeks, median cord pH = 7.01, and median 10-minute Apgar = 5. Adverse motor outcome was noted in 23 of 117 (20%). Putamen/globus pallidus injury on T1, GA, and cord pH were the most informative features. Feature selection improved model accuracy from 79% (48-feature MRI model) to 85% (three-feature model). The three-feature DL model had superior performance to the best LR model (area under the receiver-operator curve 0.69 versus 0.75). CONCLUSIONS: The parsimonious DL model predicted adverse HIE motor outcomes with 85% accuracy using only three features (putamen/globus pallidus injury on T1, GA, and cord pH) and outperformed LR.

Re-examining the arterial cord blood gas pH screening criteria in neonatal encephalopathy.

OBJECTIVE: Screening criteria for neonatal encephalopathy remain a complex combination of subjective and objective criteria. We examine the utility of universal cord blood gas testing and mandatory encephalopathy evaluation for infants with pH ≤7.10 on umbilical cord arterial blood gas (cABG) as a single screening measure for timely identification of moderate/severe encephalopathy. DESIGN, SETTING, PATIENTS: Infants born at a single centre between 2008 and 2015, who were ≥36 weeks, had no congenital anomalies and had a cABG pH ≤7.10 were identified for a retrospective cohort study. Maternal/perinatal and patient factors were collected. RESULTS: 27 028 infants were born during the study period; 412 met all inclusion criteria. Of those, 35/85 infants with pH <7.00 and 34/327 infants with pH between 7.00 and 7.10 had moderate/severe encephalopathy. Encephalopathy was identified on the basis of pH and examination alone (no other perinatal criteria present) in 5/35 and 13/34 infants in the two pH groups, respectively.A cABG pH threshold of ≤7.10 was associated with a sensitivity of 74.2% and a specificity of 98.7% for detection of moderate/severe encephalopathy. Based on these data, 25 infants with cABG pH between 7.00 and 7.10 will need to be screened to identify one neonate with moderate/severe encephalopathy, who might have otherwise been missed using conventional screening, a 15% increase in appropriate selection and treatment over current methods. CONCLUSION: Universal cord blood gas screening with a pH threshold ≤7.10 and mandatory encephalopathy examination results in greater detection of infants with moderate/severe encephalopathy and timely initiation of therapeutic hypothermia.

A validated clinical MRI injury scoring system in neonatal hypoxic-ischemic encephalopathy.

BACKGROUND: Deep nuclear gray matter injury in neonatal hypoxic-ischemic encephalopathy (HIE) is associated with worse neurodevelopmental outcomes. We previously published a qualitative MRI injury scoring system utilizing serial T1-weighted, T2-weighted and diffusion-weighted imaging (DWI), weighted for deep nuclear gray matter injury. OBJECTIVES: To establish the validity of the MRI scoring system with neurodevelopmental outcome at 18-24 months. MATERIALS AND METHODS: MRI scans from neonates with moderate to severe HIE treated with therapeutic hypothermia were evaluated. Signal abnormality was scored on T1-weighted, T2-weighted and DWI sequences and assessed using an established system in five regions: (a) subcortical: caudate nucleus, globus pallidus and putamen, thalamus and the posterior limb of the internal capsule; (b) white matter; (c) cortex, (d) cerebellum and (e) brainstem. MRI injury was graded as none, mild, moderate or severe. Inter-rater reliability was tested on a subset of scans by two independent and blinded neuroradiologists. Surviving infants underwent the Bayley Scales of Infant and Toddler Development-III (Bayley-III) at 18-24 months. Data were analyzed using univariate and multivariate linear and logistic regression. RESULTS: Fifty-seven eligible neonates underwent at least one MRI scan in the first 2 weeks of life. Mean postnatal age at scan 1 was 4±2 days in 50/57 (88%) neonates and 48/54 (89%) surviving infants underwent scan 2 at 10±2 days. In 54/57 (95%) survivors, higher MRI injury grades were significantly associated with worse outcomes in the cognitive, motor and language domains of the Bayley-III. CONCLUSION: A qualitative MRI injury scoring system weighted for deep nuclear gray matter injury is a significant predictor of neurodevelopmental outcome at 18-24 months in neonates with HIE.

A novel method for assessing cerebral autoregulation in preterm infants using transfer function analysis.

BACKGROUND: Autoregulatory dysfunction is an important contributor to brain injury in premature infants, particularly intraventricular hemorrhage (IVH). The autoregulatory system acts as a filter that dampens the systemic blood flow to follow a normal cerebral perfusion profile. METHODS: Simultaneous arterial blood pressure and cerebral near-infrared spectroscopy (NIRS) data were collected from infants born before 28 wk estimated gestational age. The resulting data were preprocessed and then divided into nonoverlapping 20-min epochs. The transfer function estimate was calculated to determine dampening ability. RESULTS: Sixty-two infants were prospectively recruited with a mean estimated gestational age of 25.4 ± 1.3 wk and birth weight of 832 ± 199 g. 67% were male, 24/62 had IVH, 17/62 received dopamine, 47/62 had antenatal steroid exposure, and 22/62 received fentanyl.Advancing estimated gestational age and birth weight z-score predicted stronger dampening while African-American race and IVH of any grade predicted weaker dampening. CONCLUSION: This preliminary report suggests an impairment in dampening ability associated with immaturity, decreased birth weight z-score, and African-American race. Decreased dampening is also associated with IVH, although these results cannot distinguish between decreased dampening as an antecedent or sequela of IVH. These observations should be studied in a larger sample.