基于DeepLabCut算法的小鼠步态分析系统建立及对衰老所致运动功能的评价  

Establishment of mice gait analysis system based on DeepLabCut algorithm to evaluate motor function of aging mice

作　　者：李至宏 盛益华 李由[2] 彭治香 曾星铫 谷新丽 田嘉怡 李思迪 LI Zhi-hong;SHENG Yi-hua;LI You;PENG Zhi-xiang;ZENG Xing-yao;GU Xin-li;TIAN Jia-yi;LI Si-di(College of Biology and Environmental Sciences,Jishou University,Jishou Hunan 416000,China;National Key Laboratory of Human Factor Engineering,China Astronaut Research and Training Center,Beijing 100094,China;Dept of Gastroenterology,Xiangxi Hospital of Ethnic Traditional Chinese Medicine,Jishou Hunan 416000,China)

机构地区：[1]吉首大学生物资源与环境科学学院,湖南吉首416000 [2]中国航天员科研训练中心,国家人因工程重点实验室,北京100094 [3]湘西自治州民族中医院消化科,湖南吉首416000

出　　处：《中国药理学通报》2024年第9期1792-1799,共8页Chinese Pharmacological Bulletin

基　　金：国家自然科学基金资助项目(No 81960786);湖南省自然基金面上项目(No 2021JJ30559,2022JJ30477);湖南省教育厅科学研究项目(No 21C0386);湖南省研究生科研创新项目(No QL20210245)。

摘　　要：目的建立一种基于DeepLabCut(DLC)算法用于评价老年小鼠运动功能的步态分析系统。方法基于深度学习技术中的DLC算法,采用跑台装置和全封闭设计,构建系统软硬件;应用本系统评价不同运动模式下衰老所致小鼠的步态差异;通过相关性分析探究体质量与体长对步态指标的影响。结果本系统实现特定步速下小鼠三维立体步态(侧面和腹平面)的同步分析,自动量化47项步态指标。应用本系统发现,步行时(15 cm·s^(-1)),相比2月龄、8月龄和15月龄小鼠体转角标准偏差下降,前肢摆动时长、膝关节(Knee)角度标准偏差、左后爪和右后爪向外角度平均值增加;15月龄小鼠还出现步频降低,步幅、双支撑总时长、Knee伸展和收缩距离增加。小跑时(20 cm·s^(-1)),15月龄小鼠无法稳定行走,相比2月龄,8月龄小鼠左后爪向外角度平均值和双支撑总时长增加。相关性分析发现,步频、步幅、前肢摆动时长、后爪向外角度平均值、双支撑总时长、Knee角度标准偏差、Knee伸展和收缩距离等指标均未受到体质量和体长变化的影响。结论基于DLC算法的小鼠步态分析系统实现了对老年小鼠步态更加敏感、准确、全面的评价,区分出老年小鼠为维持步态稳定性所表现的步态特征,并筛选出更能反映老年小鼠步态变化的行为指标。为今后更有效评估抗衰老、抗运动协调功能下降药物的药效与副作用提供方法学基础。Aim To establish a gait analysis system based on DeepLabCut(DLC)algorithm for evaluating motor function in aged mice.Methods Based on DLC algorithm in deep learning technology,treadmill device and fully closed design were used in the system,including software and hardware.This system was applied to evaluate gait characteristics of mice due to aging under different movement modes.Correlation analysis was used to explore the effects of body weight and body length on gait indicators.Results This system realized the synchronous analysis of three-dimensional gait(lateral and ventral plane)of mice at specific gait speed,and automatically quantified 47 gait indicators.Using this system,it was found that during walking(15 cm·s^(-1)),the standard deviation of body turning angle decreased,forelimb sway duration,standard deviation of knee angle,mean outward angles of left and right hind paw increased in 8 and 15 month-old mice,compared with 2-month-old mice.However,15-month-old mice showed decreased walking frequency,and increased stride width,total duration of double support,and knee extension and contraction distance.In addition,at trot(20 cm·s^(-1)),15-month-old mice were unable to walk steadily,and 8-month-old mice had increased total duration of double support and mean outward angles of left hind paw,compared with 2-month-old mice.Correlation analysis revealed that indicators like walking frequency,stride width,forelimb sway duration,total duration of double support,standard deviation of knee angle,knee extension and contraction distance,were not affected by changes in body weight and body length.Conclusions The gait analysis system based on DLC algorithm can achieve a more sensitive,accurate and comprehensive evaluation of the gait of aged mice,distinguishing the gait characteristics of aged mice to maintain gait stability,and selecting behavioral indicators that better reflect the gait changes of aged mice.It provides a methodological basis for more effective assessment of efficacy and side effects of drugs for ant

关 键 词：步态分析系统 人工智能 DeepLabCut 衰老 运动功能 步态特征 

分 类 号：R-332[医药卫生] R323.7R319R339.38R443.1