- 學位論文
無線可穿戴式早發性退化性關節炎診斷帶之分析研製
Development of a wireless wearable strap for diagnosis of early degenerative arthritis
摘要
膝關節是人體中最容易受到傷害和老化的關節，一旦產生病變，就會造成膝部疼痛，終至無法行走。因此，如何儘早診斷出初期的退化性膝關節炎並加以預防惡化，以及如何抑制退化性膝關節炎患者的疼痛並加以改善生活品質，就顯得格外重要及具價值性。 本論文改良先前之非侵入式無線多通道膝關節量測診斷系統，將穿戴式概念融入其中。該系統能夠同時量測三部位的振動訊號(包含外側髁突、髕骨中央、內側髁突)，為驗證該系統的可行性，已與義大醫院骨科部完成125例人體試驗，包括Control組(N=29)、OA組(N=96) (又分為Group 1：All，Group 2：OA-I，Group 3：OA-II，及Group 4：OA-III~Ⅴ)，並進行主動擺動、被動擺動、上下階梯、正常行走、及雙腳交互蹲立五種試驗。本論文分別以經驗模態分解法(EMD)、總體經驗模態分解法(EEMD)(分別加入雜訊20、100、300次)、及樣本熵方法(SampEn)來對膝關節振動訊號進行分析，並統計觀察所分離的特徵參數--包括本質模態函數(IMF)的平均頻率和能量百分比、及樣本熵值，是否可作為判斷受測者膝關節是否已有退化現象，並進一步期盼是否可作為定量退化程度的指標。 EMD分析結果顯示，OA各組和Control組的振動訊號之間，有多個IMFs之能量百分比存在顯著差異，例如：Groups 1、2及3之受試者在被動擺動時由外側髁突所測得振動訊號的IMF2之能量百分比(0.669±0.162%，0.659±0.149%，0.719±0.170%)均顯著大於Control組(0.590±0.100%)(all p<0.05)。EEMD分析結果顯示，OA各組和Control組的振動訊號之間，也有多個IMFs之能量百分比有顯著差異存在，例如：Groups 1、2、3、及4在上下階梯時由外側髁突所測得的振動訊號，經由EEMD加入300次雜訊運算後，所得到IMF3之能量百分比(0.431±0.109%，0.426±0.094%，0.434±0.109%，0.469±0.161%)均顯著大於Control組(0.374±0.070%)(all p<0.05)。SampEn分析結果顯示，OA各組和Control組之振動訊號值有顯著差異，例如：Groups 1、2、3、及4在被動擺動時由內側髁突所測得的振動訊號之SampEn值(0.544±0.205，0.560±0.127，0.496±0.214，0.415±0.234)均顯著小於Control值(0.622±0.164)(all p<0.05)。 本論文已完成無線穿戴式退化性關節炎量測裝置與系統，又發現可以利用經驗模態組、總體經驗模態組及樣本熵方法，來獲得振動訊號某些IMFs的能量百分比及樣本熵值，並可依據這些特徵參數，能有效判斷膝關節是否正常或已有退化現象。
並列摘要
Knee joints are the most vulnerable and the most easily aging joints in the human body. Once the knee joints are mal-functional, they may cause knee pain, and no longer support to walk. Therefore, it is particularly important and valuable to diagnose early degenerative knee arthritis and to prevent deterioration. Thus, the purpose of the thesis was to improve a previous wireless and wearable measurement system developed by our laboratory for non-invasive detection of osteoarthritis (OA) of knee joints. The system can simultaneously measure three parts of the vibroarthrographic signal (include Lateral condyle, Mid-patella, and Medial condyle). To validate its feasibility of the measurement system, the study recruited 125 subjects including 29 control participants and 96 OA patients (Group 1: all OA subjects, Group 2: OA-I, Group 3: OA-II, and Group 4: OA-III~IV) in the Department of Orthopedics, E-Da Hospital. Those subjects were asked to perform the active and the passive swings, the step-forward and –backward motions, the walking movement, and the squatting movement. The Empirical Mode Decomposition (EMD), Ensemble Empirical Mode Decomposition (EEMD), and Sample Entropy (SampEn) methods were employed to analyze the vibroarthrographic signals, and statistically observe the separated characteristic parameters. These parameters, including the mean frequency and the energy percent of intrinsic mode functions (IMF), and the calculated value of SampEn can be used as indicators of the degree of quantitative degradation in OA knee joints. In the EMD analysis, significant differences generally existed in the energy percentage of several IMFs of vibroarthrographic signals between the individual OA groups and the control group. For example, the energy percentage of IMF2 of vibroarthrographic signals recorded on the Lateral condyle during the passive swing in Groups 1, 2 and 3 (0.669 ± 0.162%, 0.659 ± 0.149%, and 0.719 ± 0.170%) is significantly greater than that (0.590±0.100%) in the control group (all p<0.05). In the EEMD analysis, we found that the energy percentage of several IMFs of vibroarthrographic signals in the individual OA groups differed from that in the control group significantly. For instance, by applying the EEMD with 300 noise-adding operations, we found that the energy percentage of IMF3 of vibroarthrographic signals recorded on the Lateral condyle during the step-forward and –backward motion in Groups 1, 2, 3 and 4 (0.431 ± 0.109%, 0.426 ±0.094%, 0.434 ± 0.109%, 0.469±0.161%) is significantly larger than that (0.374±0.070%) in the control group (all p<0.05). In the SampEn analysis, the SampEn values of the vibration signals were significantly different between the individual OA groups and the control group, especially for the vibration signals recorded on the position of the Medial condyle. For the vibration signals recorded on the Medial condyle during the passive swing, the SampEn values in the Groups 1, 2, 3 and 4 (0.544 ± 0.205, 0.560 ± 0.127, 0.496 ± 0.214, 0.415 ± 0.234) are significantly less than that (0.622±0.164) in the control group (all p<0.05). In summary, a wireless and wearable measurement system is improved in the thesis. The energy percent of several IMFs of vibroarthrographic signals by both the EMD and EEMD, and the SampEn value by the sample entropy are useful indicators that may be applied to distinguish the control from the OA knee joints.