3D gait analysis data of children with CP used for gait classification

The data results from a search of the database of the clinical motion analysis laboratory of University Hospital Pellenberg for gait analysis sessions of children with unilateral or bilateral spastic CP, aged between 3 to 18 years and GMFCS levels I, II, or III. Children with marked signs of dystonia or ataxia were excluded, but any previous treatments were allowed.Data collection is as described by Nieuwenhuys et al. (Nieuwenhuys A, Papageorgiou E, Molenaers G, Monari D, De Laet T, Desloovere K. Inter- and intrarater clinician agreement on joint motion patterns during gait in children with cerebral palsy. Dev Med Child Neurol. 2017, doi 10.1111/dmcn.13404). Briefly, the data was obtained by a standardized 3DGA measurement using optoelectronic cameras (Vicon Motion Systems, Oxford, UK), observing reflective markers, attached by clinical experts to the anatomical landmarks of the child’s lower legs, following the Plug-In-Gait marker configuration. Children were walking barefoot at a self-selected speed. The joint angles and their derivatives were obtained through the Nexus software. Additionally, the kinematic data was time-normalized to the overall gait cycle (pelvis in the sagittal (PS), coronal (PC), and transverse (PT) plane; hip in the sagittal (HS), coronal (HC), and transverse (HT) plane; and foot progression angle (FPA)) or to the stance and swing phase (knee during stance (KSTS) and during swing (KSWS) in the sagittal plane; ankle during stance (ASTS) and during swing (ASWS) in the sagittal plane) and interpolated resulting in 51 data points for each time-varying variable. All available trials were included in the study and classified according to the consensus-based joint motion patterns [6], by either one of the two involved clinical experts. For each patient, 2 to 15 trials were used per session, with an average number of four trials. Multiple sessions per patient could be included, involving sessions before and after intervention (botulinum toxin injections, selective dorsal rhizotomy or orthopaedic surgery), as well as follow-up sessions expression the natural history (275 patients had one session, 67 had two sessions and 14 patients had more than two sessions included in the database). The interval between gait analysis sessions was 2 to 3 months for botulinum toxin injections, and 1 year for selective dorsal rhizotomy and orthopaedic surgery. Including gait analysis sessions before and after interventions created a generic database that allows to develop classification algorithms that are valid for all clinically common patient conditions. This classification was used as the ground-truth when training and validating the automatic classifiers. Pathological gait patterns were classified in comparison to the averaged walking patterns of 56 typically developing children, aged between 5 and 18 years, who had no previous history of neuromotor or musculoskeletal disorders.本数据集源自佩伦贝格大学医院临床运动分析实验室数据库的检索结果，检索对象为年龄3~18岁、粗大运动功能分级系统（Gross Motor Function Classification System, GMFCS）分级为I、II或III级的单侧或双侧痉挛型脑性瘫痪（spastic CP）患儿的步态分析检测场次。伴显著肌张力障碍或共济失调体征的患儿被排除，但允许既往接受过任何治疗的患儿入组。数据采集流程参照Nieuwenhuys等人的研究方案（Nieuwenhuys A, Papageorgiou E, Molenaers G, Monari D, De Laet T, Desloovere K. 脑性瘫痪患儿步态期间关节运动模式的评定者间与评定者内临床一致性研究. 发育医学与儿童神经病学, 2017, doi: 10.1111/dmcn.13404）。简要而言，本数据通过标准化三维步态分析（3DGA）检测获取：采用英国牛津Vicon运动系统公司的光电运动捕捉摄像头，追踪由临床专家按照Plug-In-Gait标记方案粘贴于患儿小腿解剖标志点的反光标记物；患儿以自主选择的速度赤足行走，通过Nexus软件获取关节角度及其导数数据。此外，运动学数据按全步态周期（骨盆矢状面（PS）、冠状面（PC）与横断面（PT）；髋关节矢状面（HS）、冠状面（HC）与横断面（HT）；以及足前进角（FPA））或支撑相与摆动相（矢状面支撑期膝关节（KSTS）与摆动期膝关节（KSWS）；矢状面支撑期踝关节（ASTS）与摆动期踝关节（ASWS））进行时间归一化处理，并通过插值得到每个时变变量的51个数据点。本研究纳入所有可用的步态检测试次，并由两名参与研究的临床专家之一基于共识性关节运动模式标准[6]进行分类。每名患儿每次检测场次可纳入2~15个试次，平均为4个试次。每名患儿可纳入多次检测场次，包括干预前、干预后（肉毒毒素注射、选择性脊神经后根切断术或骨科手术）的检测，以及反映疾病自然病程的随访检测记录；数据库中共纳入275名患儿1次检测、67名患儿2次检测，14名患儿超过2次检测。接受肉毒毒素注射的患儿，其两次步态分析检测的间隔为2~3个月；接受选择性脊神经后根切断术与骨科手术的患儿，间隔为1年。纳入干预前后的步态分析检测记录，可构建通用型数据库，用于开发适用于所有临床常见患儿情况的分类算法。该分类结果被用作训练与验证自动分类器的金标准（ground-truth）。病理步态模式的分类参照56名年龄5~18岁、无神经运动或肌肉骨骼疾病既往史的典型发育儿童的平均行走模式进行。