Fidgety Philip and the Suggested Clinical Immobilization Test: Annotation Dataset & HMovements Automated Movement Detection Algorithm

ANNOTATION DATASET To facilitate ‘structured behavioral observations’ in clinical practice, we are interested in analyzing movement patterns, mainly voluntary and involuntary motions, during sitting. We developed a clinical protocol for analyzing characteristics of voluntary movements during sitting with the goal to capture disorder specific movement patterns and contribute to the phenotypic characterization of H-behaviors. To enhance our understanding, we conducted this phenotyping exercise independent of discipline-related boundaries and applied a pictogram guided-phenotyping language (PG-PL). Phase 1, Step 1: Annotation/analysis concept naïve research assistants (RAs) were trained. They then annotated three original Fidgety Philip cartoons and were instructed to ‘describe, but not interpret’. The dataset contains RAs' annotations of the cartoons. Phase 1, Step 2: RAs then annotated snapshots from suggested clinical immobilization tests (SCIT) to work out the distinction between ‘interpretive’ and ‘neutral, non-interpretive’ PG-PL descriptions in the analysis of snapshots from the SCIT with free-hand and then PG-PL annotations. The dataset contains RAs' free-hand and pictogram annotations of 12 SCIT snapshots. Phase 2: The goal of this phase was to apply the PG-PL to SCIT videos and to develop the first machine learning algorithm for automated movement detection. The dataset contains RAs' pictogram annotations of 1-minute long SCIT video clips. The data are available in raw and processed formats.​ HMOVEMENTS: AUTOMATED MOVEMENT DETECTION ALGORITHM The automated movement detection algorithm also is called "HMovements." This algorithm is available for download within this dataset. See the User Guide for instructions on using the algorithm. Note that the algorithm cannot be used on a Mac.标注数据集 为推动临床实践中的结构化行为观察工作，本研究聚焦于分析坐姿状态下的运动模式，主要涵盖自主运动与非自主运动两类。我们开发了一套用于分析坐姿自主运动特征的临床方案，旨在捕捉疾病特异性运动模式，助力H行为（H-behaviors）的表型特征刻画。为深化研究认知，我们突破学科边界开展本次表型分析工作，并采用图标引导型表型描述语言（PG-PL）。 第一阶段 步骤1 首先对无标注与分析经验的研究助理（RAs）开展培训。随后，研究助理对3部原版《烦躁的菲利普》（Fidgety Philip）动画短片进行标注，并被要求仅开展描述性记录，不得进行解读。本数据集包含研究助理对上述动画的标注结果。 第一阶段 步骤2 随后，研究助理对推荐临床制动测试（SCIT）的快照进行标注，以明确在分析SCIT快照时，"解读性描述"与"中性非解读性描述"两类PG-PL表述的边界：研究助理先以自由文本形式完成标注，再采用PG-PL进行标注。本数据集包含研究助理对12张SCIT快照的自由文本与图标标注结果。 第二阶段 本阶段目标为将PG-PL应用于SCIT视频，并开发首款用于自动运动检测的机器学习算法。本数据集包含研究助理对12段时长1分钟的SCIT视频片段的图标标注结果。数据集提供原始格式与处理后格式的数据文件。 --- 运动检测：自动运动检测算法 本自动运动检测算法亦名为"HMovements"，可在本数据集内下载获取。使用方法请参阅《用户指南》。请注意，该算法无法在Mac设备上运行。