Brain injury in newborns accounts for more than 9 million years lived with disability worldwide. The functional consequences of early brain injury in newborns are commonly seen months or even years after birth, delaying therapeutic interventions and leaving families with uncertainty about their child’s health status for months and years.
Classification of movement characteristics and movement patterns in spontaneous movements in young infants has in recent years been presented as the most accurate method for early identification of infants who will later develop serious motor problems and reduced motor function (like cerebral palsy). Today, characteristic of such normal and abnormal infant spontaneous movement could be objectively captured and processed by computer software from video recordings but the procedures are so far too imprecise, cumbersome and time consuming for clinical use (se også her).
Based on a database of 900 standardized video recordings at St. Olavs University Hospital of infants at risk of neurological dysfunctions from Norway, USA and India, this student assignment aims at solving the following tasks:
1) To identify the cerebral palsy-related movement of infant body segments (legs, arms, trunk and head) from the database of video recordings using computer vision and machine learning algorithms.
2) Make this procedure time effective, feasible and available for researchers within the medical field without any technical or computer vision expertise.
Hovedveileder: Heri Ramampiaro, IDI.
Biveiledere: Lars Adde, LBK, Fakultet for medisin og helsevitenskap
og Espen Ihlen, INM, Fakultet for medisin og helsevitenskap