Research On Auxiliary Diagnosis Of Developmental Coordination Disorder Based On Skeletal Sequences

Developmental coordination disorder is a common neurodevelopmental disorder in children characterized by impairments in hand-eye coordination,balance and postural control.Currently,diagnosing developmental coordination disorder in children relies on a manual assessment by medical professionals,which is time and energy-consuming and has the problem of subjective assessment results.This study proposes an automated movement assessment method based on skeletal sequences,which can greatly improve the efficiency and accuracy of assessment,reduce costs,reduce assessment errors caused by human factors,and better serve the healthy development of children.The specific research components and innovations are.(1)Constructing a video dataset of developmental coordination disorder and designing a localized and improved scoring scale assessment rule based on the MABC-2.A Blaze Pose-based lightweight posture estimation network is proposed to obtain the human skeletal sequence information in the videos,normalize the skeletal sequences using the hip centroid as the origin and the image width and height as the scale,and complete the skeletal sequence data pre-processing using linear interpolation according to the 3σ principle.(2)For the fine action coin toss,a skeletal sequence time series feature method is proposed to extract keyframes from the video.The composite score sums the action time score of the action,the keypoints similarity score of the keyframes,the standard deviation score of the right-hand key point coordinates of the keyframes and the time series complexity score of the right-hand keypoints coordinates using different weights.(3)For the static movement single leg balance,it is proposed to use the normal movement’s keypoints coordinate sequence and the test movement’s keypoints coordinate sequence for similarity calculation.The combined scores of the keypoints similarity score,the time duration score and the standard deviation score of the keypoints coordinates are weighted together using different weights.(4)For the dynamic action standing long jump,a keyframe approach based on manual feature design is proposed to extract keyframes from the video.A DTW-based time dimension alignment algorithm is used for the test action.The combined scores are weighted to sum the keypoints similarity scores of the action and the keyframes using different weights.This thesis is based on the study of diagnostic algorithms for developmental coordination disorders in children.It develops a supplementary diagnostic system to help doctors capture video data from children via web pages and obtain data for diagnostic analysis.The system can effectively save doctors’ and patients’ consultation time and alleviate insufficient medical resources.The use of the system is expected to be extended to various primary care facilities and regions to help more patients in need.