Analyzing the MRI Data from Premature Infants

Brain segmentation is crucial for providing accurate predictions to the patients due to the information which can be extracted from it, such as the shape and size of the tissue, underdeveloped or injured regions, and annual comparative results for brain growth. Segmentation of infants’ brains currently requires deep expertise and is a labor-intensive process, typically performed by an experienced radiologist, and typically only for research studies, and not for routine clinical care. Automatic segmentation methods would highly increase the number of infants with brain injuries for whom we would have segmentation data, enabling further studies of neurodevelopmental outcomes. The goal of our research is to build an automatic segmentation solution, which could segment brain MRIs for neonates, at-term infants, and older children. The model would be also able to label white matter injury, which is one of the most common and dangerous injuries in preterm infants, and can be used to predict neurodevelopmental outcomes. Using segmentation results and the provided meta-data in a form of clinical labels, we can build an outcome predicting model. The provided results from the prediction model are going to be useful for medical researchers to analyze the feature importance that impacted the result.

Faculty Supervisor:

Michael Brudno

Student:

Partner:

National University of Kyiv-Mohyla Academy

Discipline:

Computer science

Sector:

Artificial Intelligence

University:

University of Toronto

Program: