基于深度图像的ADHD儿童运动量化算法  

Movement quantification of children with attention deficit and hyperactivity disorder based on depth images

作　　者：李智倩 张茜倩 崔娅 张晗 李元媛 何凌[1] LI Zhi-qian;ZHANG Xi-qian;CUI Ya;ZHANG Han;LI Yuan-yuan;HE Ling(College of Biomedical Engineering,Sichuan University,Chengdu 610065,China;Huaxi Mental Health Center,Sichuan University,Chengdu 610041,China)

机构地区：[1]四川大学生物医学工程学院,四川成都610065 [2]四川大学华西心理卫生中心,四川成都610041

出　　处：《计算机工程与设计》2023年第1期74-83,共10页Computer Engineering and Design

基　　金：国家自然科学基金青年基金项目(81901389);四川省科技厅基金项目(2019YFS0236)。

摘　　要：为解决临床上对注意缺陷与多动障碍诊断耗时长,存在主观因素影响等问题,提出基于深度图像的多动儿童运动量化算法。针对多动儿童目标区域,提出基于噪声位置及灰度分布信息的去噪算法,用等值线提取多动儿童目标区域的边缘信息,分割出完整的多动儿童身体。提出基于CoM(center-of-mass)的运动时长特征,对多动儿童随时间的运动进行量化。实验对多动儿童的视频数据进行测试,其结果表明,该算法对多动儿童身体区域的分割正确率为82.73%~93.77%,运动量化正确率为88.37%~92.47%。In clinic,the diagnosis of ADHD is influenced by the subjective experiences of doctors and this diagnosing process is time-consuming.To solve these problems,an automatic movement quantification algorithm for children with ADHD was proposed based on the processing of depth images.A denoising algorithm based on location parameters and gray distribution information of depth pixels was proposed to process ROI area of ADHD.The isohypse was used to extract the edge information of ROI area for ADHD children.Based on these steps,the ROI area of ADHD child was segmented.Based on the segmentation results,the features representing the movements of ADHD child were proposed to quantify the movement based on CoM(center-of-mass).Experiments were conducted on the video data of ADHD children.The results show that the accuracies of the segmentation algorithm are 82.73%-93.77%.The accuracies of movement quantification are in the range of 88.37%-92.47%.

关 键 词：注意缺陷与多动障碍 运动量化 活动过动 运动深度传感器 去噪 等值线 深度图像分割 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]