Toward our long-term goal of delivering precision medicine in the treatment of neonatal hypoxic-ischemic encephalopathy (HIE), we plan to develop a methodological framework to classify HIE based on brain MRI evaluation combined with clinical variables to better predict neurological prognosis. In this proposal, we will create an MRI quantification tool to identify various types of lesions, which, combined with clinical variables, will isolate HIE subtypes and subsequent clinical phenotypes to predict prognosis. HIE is the most common cause of acquired brain injury in the neonatal period. It can result in a wide range of neurological complications that affect various functional domains, with heterogeneous severity. Stratification of HIE subtypes and specific prognoses is essential for developing and delivering targeted adjuvant and rehabilitative treatments and is also necessary for medical providers in order to guide the appropriate allocation of resources. Although predictive biomarkers have been highly anticipated, as of yet, there are none validated. MRI has demonstrated strong predictive power for severe neurobehavioral deficits within the context of severe MRI findings. However, predicting outcomes following moderate-to-mild changes or even a normal-looking brain MRI does not guarantee normal neurobehavioral outcomes. With the recent advances in image analysis technologies, we intend to increase the sensitivity and negative predictive value by detecting and quantifying moderate-to-mild pathological changes, which are difficult to evaluate qualitatively. Since individualized prediction cannot be made from a single feature, as each feature weakly correlates with outcomes, we hypothesize that patient stratification, combining brain MRI features and clinical characteristics, will be highly accurate for individualized prediction. We will apply our automated structure-by-structure image quantification (SIQ) pipeline, developed and validated through R01HD065955, to be applied for the MRI quantification in this proposal. The HIE cohort study (R01HD086058) will provide a library of teaching files that consist of MRIs with various types of lesions, from which the SIQ algorithm learns the features of the lesions. The cohort also includes clinical variables, such as serum markers and electroencephalograms, combined with the MRI features and test data for the validation study. For Aim 1, we will create a reference library that includes MRI atlases with various pathological changes due to HIE. Combined with the multi-atlas label fusion and lesion localization algorithms, the library enables a robust SIQ. For Aim 2, we will apply a supervised learning algorithm to the MRI features quantified by the SIQ to identify brain lesions and the severity that is associated with certain outcomes. Aim 3 will use a supervised classification algorithm for the MRI features and clinical variables to determine the HIE subtypes related to the affected functional domains and ...