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.

Decreasing Unplanned Extubation in the Neonatal ICU With a Focus on Endotracheal Tube Tip Position.

BACKGROUND: Unplanned extubation (UE) is an important quality metric in the neonatal ICU that is associated with hypoxia, bradycardia, and risk for airway trauma with emergent re-intubation. Initial efforts to reduce UE in our level 4 neonatal ICU included standardized securement of the endotracheal tube (ETT) and requiring multiple providers to be present for ETT adjustments and patient positioning as phase 1 interventions. After an initial decline, the UE rate plateaued; an internal retrospective review revealed that the odds of UE were 2.9 times higher in the setting of an ETT tip at or above T1 (high ETT) on chest radiograph just prior to UE. The team hypothesized that advancing ETT tips to below T1 would reduce UE risk in infants of all gestational ages. METHODS: Over a period of 32 months, we compared pre-intervention and post-intervention UE rates in our neonatal ICU after a 2-step initiative that focused initially on ETT securement and assessment, with a subsequent addition of a single intervention to advance ETT tips below T1. To determine if the decrease in UE rate could be secondary to our intervention, data were analyzed from 3 cohorts: a control group of 40 infants with 185 chest radiographs and no UEs, 46 infants with chest radiographs prior to 58 UE events before the intervention, and 37 infants with chest radiographs prior to 48 UE events following the intervention. RESULTS: Advancing ETT tips below T1, in addition to the use of a standard UE-prevention bundle, led to a significant decrease in the UE rate from 1.23 to 0.91 UEs per 100 ventilator days, with 14% of postintervention UEs attributed to ETT advancement. CONCLUSIONS: High ETTs are significantly associated with UEs in the neonatal ICU. Optimizing ETT position may be an underrecognized driver in the provider's toolbox to reduce UEs. Because ETT repositioning carries risk of UE, extra caution should be taken during advancement.