帕金森氏症是一種神經退化疾病,起因於中腦處黑質紋狀體中的多巴胺神經細胞的逐漸退化或死亡。步態的失序為帕金森氏症為主要的臨床表徵之一,會出現於輕度的帕金森患者身上。故而帕金森氏症患者的步態分析成為臨床醫師辨別患者與進行治療的重要依據。現有的步態量測技術如Vicon動作分析系統是一個高精度與信效度的步態量測工具,然而Vicon系統不易架設於一般診所或住家中,使得帕金森氏症步態分析受到極大的限制。為解決此一問題,本論文提出兩個利用電腦輔助的單一影像序列處理方法來協助分析輕度帕金森步態。 第一研究階段是用核心主成份分析法來量化與辨識穩定行走期的輕度帕金森步態。實驗環境需要布置一與患者有強烈對比色差的背景及一6公尺長的步道;實驗設備為一手持式數位單眼攝影機,從受試者側面拍攝步態啟動時與穩定行走狀態時的步行表現。結果顯示核心主成份分析法在辨別健康受試者、輕度帕金森病患在接受左多巴治療前、以及接受左多巴治療後的步態表現時,使用核心主成份分析法可以獲得高達80.51%的敏感度;使用核心主成份分析法也可以輕易取得量化的步態參數。而步態在頻譜上的能量分佈顯示接受左多巴治療前的輕度帕金森患者的步態能量頻譜主頻帶顯著低於健康受試者及接受左多巴治療後的輕度帕金森患者兩個族群的的步態能量頻譜主頻帶。量化的時域步態參數和由輕度帕金森氏症統一評分量表之三擷取出的五種動作指標分數的關連性則顯示,量化的步態參數也可以反映出輕度帕金森患者的上肢動作障礙嚴重程度。 步態啟動表現以及步態對稱性也是用來辨別與治療輕度帕金森患者的重要依據。由於第一研究階段使用的方法仍有極大的空間限制,無法在某些僅能提供有限空間的診所使用。因而在第二研究階段,中心追蹤演算法被用來量化健康受試者、輕度帕金森病患在接受左多巴治療前、以及接受左多巴治療後的步態啟動表現以及步態對稱性。希望藉由量化的步態啟動表現以及步態對稱性,能快速且準確的辨識出輕度帕金森病患,以及追蹤輕度帕金森病患的疾病進程。第二研究階段的實驗需要在受試者的腓骨頭與腳踝側面各黏貼一與背景有強烈對比的彩色標誌;實驗設備仍為一手持式數位單眼攝影機,用以從受試者側面拍攝步態啟動時與穩定行走狀態時的步行表現。結果顯示,中心追蹤演算法可以協助研究人員量化受試者的步態啟動與穩定行走狀態期間的步態表現與對稱性;量化數據則指出輕度帕金森患者在步態啟動與穩定行走期間的步態對稱性均顯著劣於健康受試者。此發現顯示只要使用步態啟動期間的步態對稱性即可辨識出輕度帕金森氏患者;此外,量化的步態表現數據亦可用於協助臨床人員區分輕度帕金森患者的病情以及了解患者接受藥物或復健治療後的改善情形。在實驗設置及演算法經過調整後,基於中心追蹤演算法的步態分析方法也可用於協助臨床醫師量化與辨識如腦部損傷或脊髓損傷等其他神經肌肉病患的步態。
Parkinson’s disease (PD) is a neurodegenerative disease of the central nervous system resulting from the death of dopamine-controlling cells in the substantia nigra within the mid-brain. Gait disorders present as cardinal motor symptoms and may be observed in early stages of the disease; therefore, the assessment of gait performance during walking has become an important reference for identification of people with PD and medical treatments. However, current available techniques for gait analysis, such as the Vicon motion analysis system with validity, reliability, and high precision, are not easily accessible in clinics or even at home for clinicians and researchers. Herein, we present two computer-aided gait analysis methods utilizing the monocular image sequences of walking to track and analyze the parkinsonian gait pattern. The first method uses kernel-based principal component analysis (KPCA) is developed to assist the recognition and quantification of mild parkinsonian gait during the steady-state walking. It requires a digital camcorder to capture the lateral view of each subject’s walking silhouette and a decorated corridor setup. The KPCA is verified to have higher sensitivity, 80.51% in this study, than the traditional image area and principal component analysis (PCA) approaches for classifying non-PD controls and “Drug-Off/On” mild PD patients. Quantitative gait parameters are obtained and the power spectrums of the patient’ gaits are analyzed. It is found that the Drug-Off mild PD patients show a lower main power spectral frequency than those of the non-PD controls and Drug-On mild PD patients. In addition, the correlations between five subscores based on the unified Parkinson’s disease rating scale (UPDRS) part III motor scores and the extracted kinematic gait parameters are discussed. Results show the feasibility of using gait performance to evaluate the motor function of mild PD patient’s upper extremity. The disordered gait initiation and walking asymmetry are instructors of postural instability and represent the major symptoms in the early stage of this disease except the affected gait performance. To provide the recognition and quantification of mild parkinsonian gait in clinics with too small space to deploy a corridor. The second method using the centroid tracking algorithm (CTA) is developed to quantify each subject’s gait parameters and the associated symmetry indexes during the gait initiation and steady-state walking periods. This method requires only a digital camcorder to capture the lateral view of each subject’s walking and two identifiers respectively secured at two palpable anatomic landmarks, the fibula head and lateral malleolus. The second method therefore becomes easier to install and use in clinics than the first one. Results in this study indicate that the method using the CTA can help clinicians and researchers quantify the gait performance and associated symmetry indexes among age-matched non-PD controls and mild PD patients in different states. Quantitative analysis reveal that the age-matched non-PD controls presented superior gait performance and associated symmetry indexes to the mild PD patients in the different states. The findings in this study suggest that the recognition of mild PD patients can be easily attained using only the gait symmetry indexes during the gait initiation period, indicating the cost and effort for diagnosis of PD patients in the early stages can be reduced largely. Besides, the quantification of gait performance may assist the clinicians to rate the severity of and monitor the progression of PD, and evaluate the therapeutic effect brought about by drug management or rehabilitation programs. In addition to performing gait analysis for patients with mild PD, we believe that the proposed portable system has the potential to help clinicians and researchers assess the gait performance of patients with other neuromuscular issues, such as traumatic brain injury and spinal cord injury.