帧间差分法在婴儿全身运动视频质量评估中的应用

Application of inter-frame difference method in evaluation of video quality for infant general movements

摘要

目的 探讨一种基于帧间差分法的计算机视频质量检测方法,以客观、高效地识别并剔除不适合进行全身运动质量评估(GMs)的低质量婴儿运动视频。方法 本研究采用分层随机抽样和简单随机抽样,选取2024年6月至2025年5月在西安交通大学第一附属医院新生儿科临床场景采集或者家庭场景采集的婴儿GMs评估视频8例(医院临床采集6例,家庭采集2例),以整段视频平均运动量作为核心质量的量化指标,并引入图像后处理算法以排除环境干扰,最后通过阈值判断视频是否合格。结果 利用8例婴儿GMs视频的前1000帧构成数据集进行验证,该方法能够有效区分婴儿活跃运动状态与静止/睡眠状态,处理速度达到平均64.60帧/秒,具备实时检测能力。通过设定合理的阈值,可自动筛选出因婴儿无运动而导致评估价值极低的视频。结论 本研究成功将帧间差分法这一技术应用于婴儿运动视频的质控环节,有望提升全身运动评估前视频筛选的效率和标准化水平,为基于视频的早期神经发育筛查提供可靠的数据质量保障。

Abstract

Objective To develop a computer-based video quality assessment method using inter-frame difference algorithm so as to objectively and efficiently screen and exclude low-quality infant movement videos unsuitable for General Movements Assessment (GMs). Methods Eight infant GMs videos (6 videos from hospital clinical acquisition,2 videos from home environments acquisition) via stratified and simple random sampling method acquired over a period from June 2024 to May 2025 were selected. The average quantity of movement (AQoM) obtained from each full video served as quantitative metric for core quality. An image post-processing algorithm was applied to reduce environmental interference,and an optimized threshold was applied to determine video eligibility. Results Through validation for dataset that consisted of the first 1000 frames from each of the eight videos,the computer-based video quality assessment and screening method effectively differentiated infant􀆳s active movement states from inactive/sleep states. The average processing speed reached to 64.60 frames per second,wihich could support real-time application. By using an optimized threshold,this video quality assessment and screening method successfully identified out and eliminated those videos with minimal utility value when the baby was not moving. Conclusion The inter-frame difference method is used successfully in quality control of infant movement videos in this study. This method is expected to improve efficiency and standardization of video pre-screening before GMs,thereby provide reliable data quality assurance for video-based early neurodevelopmental screening of the infant.

关键词

婴儿 / 全身运动质量评估(GMs) / 脑性瘫痪 / 图像处理 / 视频质量评估 / 帧间差分法

Key words

infant / General Movements Assessment (GMs) / cerebral palsy / image processing / video quality assessment / inter-frame difference method

参考文献

[1] 张建奎, 李晓捷, 唐久来, 等. 中国脑性瘫痪康复指南(2022)核心内容解读[J]. 中华实用儿科临床杂志, 2022,37(24): 1841-1853.

[2] 张娜, 崔欣华, 王立云, 等. 依托扭动运动阶段全身运动质量评估对高危儿进行个性化早期综合干预的效果分析[J]. 中国妇幼健康研究, 2024,35(7): 44-49.

[3] John H B, Oommen S P, Swathi T O, et al. Preterm general movements in prediction of neurodevelopmental disability and cerebral palsy at two years:a prospective cohort study[J]. Indian Pediatr, 2022,59(10): 769-773.

[4] 宋文亚, 倪育, 罗翠仪, 等. 全身运动评估与婴儿运动表现测试的相关性研究[J]. 中国妇幼健康研究, 2022,33(2): 1-4.

[5] Barnes F, Graham L, Loganathan P, et al. General movement assessment predicts neuro-developmental outcome in very low birth weight infants at two years:a five-year observational study[J]. Indian J Pediatr, 2021,88(1): 28-33.

[6] Zhussupova Z, Jaxybayeva A, Ayaganov D, et al. General movement assessment efficacy for assessment of nervous system integrity in children after hypoxic-ischemic encephalopathy in middle income countries[J]. Early Hum Dev, 2024,192: 105992.

[7] 蔡荣兰, 蔡娟, 杨丽娟. 线上全身运动评估在区域化高危儿随访管理中的应用研究[J]. 赣南医科大学学报, 2024,44(12): 1284-1288.

[8] Morais R, Le V, Morgan C, et al. Robust and interpretable general movement assessment using fidgety movement detection[J]. IEEE J Biomed Health Inform, 2023,27(10): 5042-5053.

[9] Valle S C, Støen R, Sæther R, et al. Test-retest reliability of computer-based video analysis of general movements in healthy term-born infants[J]. Early Hum Dev, 2015,91(10): 555-558.

[10] Adde L, Helbostad J L, Jensenius A R, et al. Early prediction of cerebral palsy by computer-based video analysis of general movements:a feasibility study[J]. Dev Med Child Neurol, 2010,52(8): 773-778.

[11] Sun Y, Zhi X, Han H. Enhancing UAV detection in surveillance camera videos through spatiotemporal information and optical flow[J]. Sensors (Basel), 2023,23(13): 6037.

[12] Cantrell D R, Cho L, Zhou C, Background subtraction angiography with deep learning using multi-frame spatiotemporal angiographic input[J]. J Imaging Inform Med, 2024,37(1): 134-144.

[13] Nie G Y, Bodda S S, Sandhu H K, et al. Computer-vision-based vibration tracking using a digital camera:a sparse-optical-flow-based target tracking method[J]. Sensors (Basel), 2022,22(18): 6869.

[14] 李源. 基于高斯建模法与背景减除法的视频移动侦测性能对比分析[J]. 集成电路应用, 2025,42(1): 10-12.

[15] Yildirim C, Asaliog ˇlu A, Coşkun Y, et al. General movements assessment and Alberta Infant Motor Scale in neurodevelopmental outcome of preterm infants[J]. PediatrNeonatol, 2022,63(5): 535-541.

[16] 中国康复医学会儿童康复专业委员会, 中国残疾人康复协会小儿脑性瘫痪康复专业委员会, 中国医师协会康复医师分会儿童康复专业委员会, 等. 中国脑性瘫痪康复指南(2022)第二章:脑性瘫痪高危儿的评定与干预[J]. 中华实用儿科临床杂志, 2022,37(13): 974-982.

[17] Marschik P B, Kwong A K L, Silva N, et al. Mobile Solutions for clinical surveillance and evaluation in infancy-general movement apps[J]. J Clin Med, 2023,12(10): 3576.

[18] 郭莛锐. Kinect v2深度相机在立体足迹三维扫描提取中的应用研究[J]. 机电产品开发与创新, 2025,38(4): 100-102.

[19] 张世宇, 祁子禹, 干智超, 等. 基于Kinect v2低成本动作捕捉系统在步态分析中的开发及应用[J]. 中国神经精神疾病杂志, 2023,49(1): 54-59.

[20] 汪志成, 赵杰, 沈博韬, 等. 基于机器视觉的婴儿全身运动质量智能评估[J]. 科学技术与工程, 2023,23(33): 14278-14286.

[21] 黄昭润. 基于动作捕捉的婴幼儿动作模式识别方法研究[D]. 沈阳: 沈阳工业大学, 2024.

[22] 张永强, 宋美霖, 刘天虎, 等. 混合骨骼特征的三帧间差分手势识别方法研究[J]. 智能科学与技术学报, 2022,4(4): 592-599.

[23] Shi L, Lu Y, Dvornek N, et al. Automatic inter-frame patient motion correction for dynamic cardiac pet using deep learning[J]. IEEE Trans Med Imaging, 2021,40(12): 3293-3304.

基金资助

国家自然科学基金(62402375)