Quantification of hand jerk asymmetry for infants in hemiplegic cerebral palsy assessment

Sammanfattning: Hemiplegic cerebral palsy (HCP) is the most common form of cerebral palsy (CP) and affects roughly 1 in 400 infants. Unable to properly use the arm and leg of one side of the body, the disorder can have major negative impacts on a person’s life. New and improved methods of treating HCP have emerged in recent decades. However, detecting high risk of developing HCP early in infants is a time-consuming and expensive process, and it is prone to human error. The assessment includes many variables and among them is the aspect of shaky and unstable movements in either arm. This factor is rated vaguely based on an expert’s experience. The rating of shakiness can vary depending on the expert that is doing the examination. In our study we show that there is some possibility to be able to quantify the shakiness and instability in arm movements. Investigation of the risk of developing HCP is usually done by examining a 5-10 minute video of a young infant playing and interacting with both arms. We used a machine learning application to determine the position of both hands in relation to the rest of the body for each frame in these types of 5-10 minute videos. We then extracted velocity, acceleration as well as jerk, which is the third derivative of position, from the position of the hands throughout the video. We compared the data of different infants and looked for differences in the jerks of their hand movements and compared it to how experts had evaluated the children based on the videos. We also performed statistical tests on the jerks to see if the right or left hand had any noticeable differences. Our hypothesis was that the shakiness and instability of the movements of the arms would correspond to rapid changes in acceleration, which would be visible in the distribution of jerks in the affected arm. Asymmetry in the jerk between the arms could thus indicate a difference in the stability and control of the arms. We have developed a method to investigate asymmetry in jerk between the hands, but due to low sample size and the lack of complete information of the diagnosis for each child, it is hard to compare our metrics against the evaluations of the children. Preliminary results show that children with great evaluations and no prenatal stroke had lower amounts of asymmetry in jerk compared to the children with worse evaluation and the ones who had been affected by a prenatal stroke. Quantifying jerk asymmetry in the arm movement of infants would allow for the avoidance of human error and make the rating assessment more concrete and efficient. Our work could thus decrease the possibility of a child not receiving a correct diagnosis and possibly miss out on potential treatment. This would promote good health well being, but also help reduce medical inequality as the assessment process becomes more standardised and definite. Our study is a step towards automatizing parts of the assessment process and achieving accurate and reliable automatization of the entire assessment process would have major impacts on HCP assessment. It could also help to shine light on the potential of quantifying and automatizing parts of other areas av cerebral palsy assessment, such as General Movements Assessment (GMA).