基於神經發展治療及深度學習之移動式自動化步態訓練機研發

The Development of a Movable Automatic Gait Rehabilitation Device based on Neuro-Developmental Treatment and Deep Learning

李友棋(You-Chi Li)

指導教授 ： 王富正

國立臺灣大學/工學院/機械工程學研究所/碩士(2020年)

https://doi.org/10.6342/NTU202001968

摘要

本論文主要針對中風患者行走能力的復健，接續先前開發的移動式神經發展治療(Neuro Developmental Treatment；NDT)步態訓練系統，並透過深度學習的方式發展一套步態偵測系統。治療手法上設計了一套可以提升受測者腰部旋轉幅度的訓練方法，並以臨床實驗驗證其效果。
NDT步態訓練整合了各種支撐、誘發、引導、與抑制等手法，是相當有效的中風復健方式，然而對於治療師而言，傳統的NDT治療方式往往相當耗時、費力，無法給予患者大量的治療復健，而且僅能在特定之空間範圍內進行復健，故本實驗室在過去開發了一套移動式自動化NDT步態訓練機，用來降低治療師的工作負擔並同時增加患者的復健訓練時間，以提升NDT步態訓練之效果。本論文延續移動式NDT步態訓練機架構，開發一套步態偵測系統及一套可以提升腰部旋轉的訓練方式。
首先，本論文提出了一套穿戴式步態偵測系統，運用深度學習即時偵測特定的步態相位，作為系統介入時機點的依據。其次，根據先前研究結果顯示，自動化NDT步態訓練機對增進縱向擺動期對稱性的效果顯著，但是在腰部旋轉角度的表現上則不如治療師之訓練，因此本論文進一步分析治療師之施力手法與腰部轉角之關係，並提出了一套可以增進腰部轉角表現的控制流程，並將其與移動式NDT步態訓練機整合，提升NDT步態訓練系統的性能。本論文運用系統識別的方式來建立數學模型，接著針對此系統模型設計控制器。最後則以擬中風受測者的NDT步態訓練，驗證所提出的訓練方法可以有效提升受測者的訓練效果。

關鍵字

深度學習 ； 神經發展治療 ； 步態訓練 ； 中風 ； 強韌控制 ； 馬達控制

並列摘要

This thesis proposes a gait detection method that can identify specific gait features in real-time and develops a training method that can help improve their pelvic rotation during neural development treatment(NDT) rehabilitation.
NDT is an effective rehabilitation method for stroke patients, because it can let patients have the feeling on walking with minimal physical intervention. However, it is very labor-intensive and time-consuming for therapists. Consequently, patients usually cannot receive sufficient training because of the shortage of therapist assistance. Therefore, we developed a movable automatic gait trainer to relief therapists’ working load and increase patients’ training time in previous research. This thesis extends the work by proposing an online gait detection algorithm and developing a new training method to improve the user’s pelvic rotation
The first contribution of this thesis is the development of a gait detection system. Gait detection and analyses are importance in clinical applications because they can help speeding up diagnosis and applying suitable treatments. In NDT rehabilitation, we need to identify specific gait events to trigger corresponding physical intervention. However, the identification of gait events for stroke patients might be difficult because of the influences of diseases. Therefore, we develop a gait detection system that can correctly identify the gait features in real-time. First, we apply a wearable device to receive subjects’ gait information. Second, we propose a recurrent neural network to build a gait detection system that can automatically extract gait features. Last, we integrate this system with the movable NDT gait trainer to conduct real-time gait detection for triggering rehabilitation intervention.
The second contribution of this thesis is the development of an intervention method to improve the user’s pelvic rotation during NDT rehabilitation. Previous studies showed that the NDT trainer was effective in improving the symmetry of swing phases, but the amplitude of pelvic rotation was not improved when comparing with the therapist guiding. Therefore, we propose a new training method to improve the pelvic rotation. First, we record the therapist’s actions and the test subject’s motions during the training. Then we translate the clinical know-how to build an expert dataset and propose an effective method to improve the pelvic rotation. Second, we apply identification techniques to derive the transfer function of the motor systems and design robust controllers to cope with disturbances and noises during gait training. Last, we recruit fourteen subjects to conduct NDT gait training to verify the proposed method. Based on the results, the proposed method is deemed effective in improving of swing phases and the pelvic rotation simultaneously during NDT rehabilitation.

並列關鍵字

deep learning ； recurrent neural network ； NDT ； gait training ； stroke ； robust control ； motor control

參考文獻

[1] World Health Organization. (2018). The top 10 causes of death. Available：https://www.who.int/news-room/fact-sheets/detail/the-top-10-causes-of-death

[2] Centers-for-Disease-control-and-prevention. (2017). Stroke Facts. Available: https://www.cdc.gov/stroke/facts.htm

[3] Go, A. S., Mozaffarian, D., Roger, V. L., Benjamin, E. J., Berry, J. D., Blaha, M. J., ... Fullerton, H. J. (2014). Executive summary: heart disease and stroke statistics—2014 update: a report from the American Heart Association. Circulation, 129(3), pp.399-410.

[4] Hsu M. H. (1996).Outcome of Rehabilitation of stroke patients in recent 10 years. Chinese J Rehabil, 24(1), pp.41-49.

[5] K. Bobath and B. Bobath. (1984). The neuro-developmental treatment., Management of the motor disorders of children with cerebral palsy. Clinics in Developmental Medicine, 90, pp.6-18.