Observation and examination of the infant's spontaneous movements at 3 months after term, using the General Movement Assessment method, has proven to be a very good method for early prediction of CP. Using video recordings of the infant's movements at 10-18 weeks after term, a trained expert can identify movement patterns that are associated with later diagnosis of CP. Physiotherapists and pediatricians at Rehabiliteringsklinikken and Barne og ungdomsklinikken at St. Olavs Hospital have acquired this expert knowledge.

Computer based motion analysis using artificial intelligence

Observation of infant movement patterns from video, using the computerized motion analysis method (GMA), requires trained and educated experts. This is expensive and relies on highly skilled personnel. Therefore, most sick newborns that require such examinations will not have access to this expertise internationally.

The video footage collected in the national and international multicenter studies using the GMA method for early prediction of CP, is used by our researchers to develop an automatic computer-based motion analysis. We have been working on this for several years, but the work has accelerated since modern methods for training computer-based algorithms by artificial intelligence (AI) have become available in recent years.

We have developed an automatic 'tracker' that uses AI methods to track the position of different parts of the child's body from the video footage.

These positions provide data that is used to train a computer to recognize early movements typical of children with and without CP. This new computer-based method is still under development, but shows very promising results where the risk of later CP is predicted with high accuracy - automatically, without the need for highly skilled experts. If you want to read more about artificial intelligence, you can visit the NTNU Open AI-Lab website and watch a video about AI and infant motion analysis. 

Read our latest research article on a major international clinical trial: Development and Validation of a Deep Learning Method to Predict Cerebral Palsy From Spontaneous Movements in Infants at High Risk, Jama Network Open, 2022