当前位置: X-MOL 学术Multimed. Tools Appl. › 论文详情
Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)
Objective and automatic assessment of bradykinesia in Parkinson’s patients using new repetitive pointing task with machine learning approach
Multimedia Tools and Applications ( IF 3.0 ) Pub Date : 2024-03-07 , DOI: 10.1007/s11042-024-18400-y
Jinee Goyal , Padmavati Khandnor , Trilok Chand Aseri

Bradykinesia is one of the main symptoms to diagnose Parkinson’s disease (PD). It indicates the slowness in the movement and can be measured using various upper-limb and lower-limb activities based on Unified Parkinson’s Disease Rating Scale. The objective methods using machine learning techniques can provide an objective assessment of bradykinesia which solves the problem of inter-rater inconsistency. In this paper, we have proposed a new Repetitive Pointing Task (RPT activity) to assess upper-limb bradykinesia. The data is taken from 17 PD + 9 healthy controls who performs RPT activity and data is collected using a 3D motion capture system. New features are calculated based on characteristics of bradykinesia and RPT activity. With only six features, our method achieved an accuracy of 97.14%, the sensitivity of 97%, the specificity of 95%, the precision of 97%, and the F1-score of 97%. We have also found the relation between RPT based features and stage estimation based on the Hoehn and Yahr scale. When compared with state-of-the-art upper limb exercises, the proposed method performs significantly better in terms of accuracy i.e., increment of 1.791% with hand grasping, 4.14% with finger tapping, 6.34% with hand clasping and 8.84% with pronation/supination activity. There is also improvement in terms of effort and time required to perform the activity as it is a simple activity that can estimate bradykinesia on both sides of the body in one single sitting of just 17.5s. There is also improvement in terms of reliability of data collection process due to its non-invasiveness. The proposed method can also estimate PD stages without additional parameters.



中文翻译:

使用机器学习方法的新重复指向任务客观自动评估帕金森病患者的运动迟缓

运动迟缓是诊断帕金森病(PD)的主要症状之一。它表示运动的缓慢程度,可以根据统一帕金森病评定量表使用各种上肢和下肢活动来测量。使用机器学习技术的客观方法可以提供运动迟缓的客观评估,解决了评估者之间不一致的问题。在本文中,我们提出了一种新的重复指向任务(RPT 活动)来评估上肢运动迟缓。数据取自 17 位 PD + 9 位执行 RPT 活动的健康对照,并使用 3D 运动捕捉系统收集数据。新特征是根据运动迟缓和 RPT 活动的特征计算的。仅用六个特征,我们的方法就达到了 97.14% 的准确度、97% 的灵敏度、95% 的特异性、97% 的精确度和 97% 的 F1 分数。我们还发现了基于 RPT 的特征与基于 Hoehn 和 Yahr 量表的阶段估计之间的关系。与最先进的上肢练习相比,所提出的方法在准确性方面表现明显更好,即抓握增量为 1.791%,手指敲击增量为 4.14%,握手增量为 6.34%,内旋增量为 8.84% /旋后活动。进行该活动所需的精力和时间也有所改善,因为这是一项简单的活动,只需 17.5 秒即可估计身体两侧的运动迟缓。由于其非侵入性,数据收集过程的可靠性也得到了提高。该方法还可以在无需额外参数的情况下估计 PD 阶段。

更新日期:2024-03-07
down
wechat
bug