Status: Funded - Closed
Quantitative MRI characterization of neonatal brain injury
Yangming Ou, PhD
BACKGROUND: Neonatal Brain Injury, specifically Hypoxic Ischemic Encephalopathy (HIE), occurs to 2‐6 per 1000 newborns in the USA, causing devastating and lasting motor, lingual, visual and intellectual function deficits.
GAP: Visual assessment of HIE in diffusion MRI based Apparent Diffusion Coefficient (ADC) maps, the current clinical routine, is subject to 20‐50% intra‐/inter‐reader variability even for experienced pediatric neuroradiologists.
HYPOTHESES: 1. Our recently developed, first of its kind, normative ADC atlases quantifying region-specific normal ADC variations are sensitive and specific to identify HIE as outliers to normal in a new patient’s ADC map; 2. Spatial patterns of the detected HIE (shape, geometry, size, ADC histogram and heterogeneity, etc.) can augment the clinically known prognostic biomarker (the locations of HIE), to better correlate with outcomes (death, disability; available to us via clinical reports).
METHODS: Study Design - In a retrospective study, we will use machine learning tools to (1) automatically label every voxel/region in a new patient’s ADC maps as HIE free or HIE affected; and (2) automatically identify spatial patterns that are better indicative of outcomes. Study Population - Clinically acquired ADC maps of N=17 HIE patients and N=14 age-matched normal controls scanned during 2006-2013 in MGH.
RESULTS: Our neonatal normative ADC atlases is freely and publicly available at https://www.nitrc.org/projects/mgh_adcatlases.
IMPACT: Research Impact- MRI characterization of HIE will boost research in understanding of the causes of HIE, in optimizing therapeutic interventions and in optimizing neuro-protection. Clinical Impact - Our ongoing close collaborations with pediatric neuroradiologists at the Massachusetts General Hospital will help deliver our atlases and atlas-based HIE characterization software tools to both radiology reading rooms and resident classrooms in MGH.