本研究為前瞻性研究(prospective study)，目的在於利用復健前測量腦電圖的資料以及機器學習技術預測復健後是否會預測的準確。在2012年到2013年與台北榮總醫院的復建中心合作，在此期間本研究收集37位中風患者，而每位患者經過24小時的復健訓練以及經由三種臨床量表(FM、TEMPA、WMFT)來評估中風患者的復健情況。
本研究採用監督機器學習的方法，將患者分為復健良好(good)以及復健一般(general)兩種情況，如何定義患者復健良好或復健一般，本研究根據臨床量表(FM、TEMPA、WMFT) 定義兩種情況：一種以三種臨床量表復健後進步分數的10%作為是否會復健良好的依據(即 ∆FM> =6.6 ; ∆WMFT >=8.5; ∆TEMPA >=13.8 )，稱之為TypeI，結果有20位中風患者屬於復原良好的；另一種以三種臨床量表復健後的分數是否達到總分之60%作為是否會復健良好的依據(即 FM >=40 ; WMFT >=45, TEMPA> =-55 )，稱之為TypeII，結果有23位中風患者屬於復原良好的。
在復健療程前，患者做80次上舉動作同時收取腦電圖，而後將腦電圖資料經過前處理(濾波、切段)後，利用廣義逆電場矩陣將前處理過的腦電圖資料推估出對側初級運動皮質區(CM1)、同側初級運動皮質區(IM1)、對側前運動區(CPM)、同側前運動區(IPM)以及輔助運動區(SMA)等五種運動區在大腦上的近似位址，再來把五種運動區所計算出的光譜密度圖利用莫萊小波(wavelet number: 7)轉換成時間頻譜圖。每次舉手所收取的腦電圖經過上述的處理後所得到的時頻圖絕對值平均起來而後進入動態因果模型之誘發響應當作該模型的觀察資料。經由動態因果模型所得到的參數以及五種運動區所得到光譜密度圖擔任資料的特徵，而後將這些特徵在二分類法下利用包裝法選取特徵，分別使用四種不同的分類器，支持向量機、邏輯回歸法、貝氏分類器、J48
分類結果在TypeI以及動態因果模型特徵下，在β+γ頻帶組合使用邏輯回歸法最高準確率92.95%。此外在TypeI以及動態因果模型特徵下，只用β頻帶組合使用邏輯回歸法準確率83.19%，暗示著在大腦運動網路上β律動對於其復健成效有顯著的影響。我們相信所發現的復健結果所帶來的知識可幫助研發最佳復健策略。

This study is a prospective study, aiming at accurate prediction of the rehabilitation outcome after stroke by using the pre-rehab electroencephalogram (EEG) and machine learning technique. 37 stroke patients, who admitted to the Rehabilitation Center at Taipei Veterans General Hospital from January 2012 to December 2013, were recruited for this study. All patients underwent 24-hour rehab program and the rehabilitation outcomes were measured with FM, TEMP and WMFT. For supervised machine learning methods, we first divided the data into two groups : good and general recovery, according to two criteria - (1) Type I : a level of 10% improvement of any above mentioned measures (i.e. ∆FM> =6.6 ; ∆WMFT >=8.5; ∆TEMPA >=13.8 ) after rehab was considered as good recovery as suggested in several studies, resulting in 20 good recovery patients (out of 37); and (2) Type II: the scale after rehab up to 60 % of full scales of any above mentioned measures (i.e. FM >=40 ; WMFT >=45, TEMPA> =-55 ) was labeled as good recovery, resulting in 23 good recovery patients. The EEG data were acquired during the shoulder flexion for eighty trials before the rehabilitation and were pre-processed offline for filtering and epoching. The spectral density from 4-48Hz at contralesional primary motor cortex (CM1), ipsilateral primary motor cortex (IM1), contralesional premotor area (CPM), ipsilateral premotor area (IPM) and the supplementary motor area (SMA) were obtained by projecting the EEG data to the chosen sources using the generalized inverse of the lead-field matrix over peri-stimulus time and then used a time-frequency Morlet wavelet transform (wavelet number: 7). The absolute value of the resulting time-frequency responses were averaged over trials and entered dynamic causal modeling for induced responses (DCM_IR) as the observations that the model is trying to explain. Both the spectral density at all sources and the parameters given by DCM were served as the data features. These features entered Wrapper method to select features, and the selected features went into the four different classifiers：SVM, Logistic Regression, NaiveBayes, J48 for two-class classification.
The classification result suggests that, the best accuracy rate was 92.95 % when using DCM features of β+γ frequencies of Type I data partition and Logistic Regression. Furthermore, the classification accuracy rate was up to 83.19 % when using only β frequency DCM features, indicating that beta rhythm within the motor network have a significant impact for recovery. We believe that our finding can help to facilitate the result of rehab by developing a knowledge-based rehab program.

1. Talelli, P., et al., Theta Burst Stimulation in the Rehabilitation of the Upper Limb A Semirandomized, Placebo-Controlled Trial in Chronic Stroke Patients. Neurorehabilitation and neural repair, 2012. 26(8): p. 976-987.
2. Sheorajpanday, R.V., et al., Quantitative EEG in ischemic stroke: Correlation with functional status after 6months. Clinical Neurophysiology, 2011. 122(5): p. 874-883.
3. Leon-Carrion, J., et al., Delta–alpha ratio correlates with level of recovery after neurorehabilitation in patients with acquired brain injury. Clinical Neurophysiology, 2009. 120(6): p. 1039-1045.
4. Denti, L., M. Agosti, and M. Franceschini, Outcome predictors of rehabilitation for first stroke in the elderly. European journal of physical and rehabilitation medicine, 2008. 44(1): p. 3-11.
5. Richiardi, J., et al., Classifying minimally disabled multiple sclerosis patients from resting state functional connectivity. Neuroimage, 2012. 62(3): p. 2021-2033.
6. Shen, H., et al., Discriminative analysis of resting-state functional connectivity patterns of schizophrenia using low dimensional embedding of fMRI. Neuroimage, 2010. 49(4): p. 3110-3121.
7. Wee, C.-Y., et al., Identification of MCI individuals using structural and functional connectivity networks. Neuroimage, 2012. 59(3): p. 2045-2056.
8. Zeng, L.-L., et al., Identifying major depression using whole-brain functional connectivity: a multivariate pattern analysis. Brain, 2012: p. aws059.
9. Rehme, A., et al., Identifying Neuroimaging Markers of Motor Disability in Acute Stroke by Machine Learning Techniques. Cerebral Cortex, 2014: p. bhu100.
10. Rehme, A.K. and C. Grefkes, Cerebral network disorders after stroke: evidence from imaging-based connectivity analyses of active and resting brain states in humans. The Journal of physiology, 2013. 591(1): p. 17-31.
11. Finnigan, S.P., et al., Quantitative EEG indices of sub-acute ischaemic stroke correlate with clinical outcomes. Clinical neurophysiology, 2007. 118(11): p. 2525-2532.
12. Sheorajpanday, R.V., et al., Additional value of quantitative EEG in acute anterior circulation syndrome of presumed ischemic origin. Clinical Neurophysiology, 2010. 121(10): p. 1719-1725.
13. Sainio, K., et al., Visual and spectral EEG analysis in the evaluation of the outcome in patients with ischemic brain infarction. Electroencephalography and clinical neurophysiology, 1983. 56(2): p. 117-124.
14. Cillessen, J., et al., Electroencephalography improves the prediction of functional outcome in the acute stage of cerebral ischemia. Stroke, 1994. 25(10): p. 1968-1972.
15. Szelies, B., et al., Prognostic relevance of quantitative topographical EEG in patients with poststroke aphasia. Brain and language, 2002. 82(1): p. 87-94.
16. Cuspineda, E., et al., Predicting outcome in acute stroke: a comparison between QEEG and the Canadian Neurological Scale. Clinical EEG (electroencephalography), 2003. 34(1): p. 1-4.
17. Finnigan, S.P., et al., Correlation of Quantitative EEG in Acute Ischemic Stroke With 30-Day NIHSS Score Comparison With Diffusion and Perfusion MRI. Stroke, 2004. 35(4): p. 899-903.
18. van Putten, M.J., et al., A brain symmetry index (BSI) for online EEG monitoring in carotid endarterectomy. Clinical neurophysiology, 2004. 115(5): p. 1189-1194.
19. Nagata, K., et al., Electroencephalographic correlates of blood flow and oxygen metabolism provided by positron emission tomography in patients with cerebral infarction. Electroencephalography and clinical neurophysiology, 1989. 72(1): p. 16-30.
20. Cuspineda, E., et al., QEEG prognostic value in acute stroke. Clinical EEG and neuroscience, 2007. 38(3): p. 155-160.
21. Burghaus, L., et al., Early electroencephalography in acute ischemic stroke: Prediction of a malignant course? Clinical neurology and neurosurgery, 2007. 109(1): p. 45-49.
22. Tecchio, F., et al., Outcome prediction in acute monohemispheric stroke via magnetoencephalography. Journal of neurology, 2007. 254(3): p. 296-305.
23. Sheorajpanday, R.V., et al., Quantitative EEG in ischemic stroke: correlation with infarct volume and functional status in posterior circulation and lacunar syndromes. Clinical Neurophysiology, 2011. 122(5): p. 884-890.
24. Friston, K., et al., Functional connectivity: the principal-component analysis of large (PET) data sets. Journal of cerebral blood flow and metabolism, 1993. 13: p. 5-5.
25. Brown, R. and L. Kocarev, A unifying definition of synchronization for dynamical systems. Chaos: An Interdisciplinary Journal of Nonlinear Science, 2000. 10(2): p. 344-349.
26. Chen, C., S.J. Kiebel, and K.J. Friston, Dynamic causal modelling of induced responses. NeuroImage, 2008. 41(4): p. 1293-1312.
27. Lisman, J.E., Relating hippocampal circuitry to function: recall of memory sequences by reciprocal dentate–CA3 interactions. Neuron, 1999. 22(2): p. 233-242.
28. David, O., J.M. Kilner, and K.J. Friston, Mechanisms of evoked and induced responses in MEG/EEG. Neuroimage, 2006. 31(4): p. 1580-1591.
29. Chen, C.-C., et al., A dynamic causal model for evoked and induced responses. NeuroImage, 2012. 59(1): p. 340-348.
30. Derdikman, D., et al., Imaging spatiotemporal dynamics of surround inhibition in the barrels somatosensory cortex. The Journal of neuroscience, 2003. 23(8): p. 3100-3105.
31. Okun, M. and I. Lampl, Instantaneous correlation of excitation and inhibition during ongoing and sensory-evoked activities. Nature neuroscience, 2008. 11(5): p. 535-537.
32. Bruno, R.M. and D.J. Simons, Feedforward mechanisms of excitatory and inhibitory cortical receptive fields. The Journal of neuroscience, 2002. 22(24): p. 10966-10975.
33. Sun, Q., The Missing Piece in the``Use It or Lose It''Puzzle: Is Inhibition Regulated by Activity or Does it Act on its Own Accord? Reviews in the Neurosciences, 2007. 18(3/4): p. 295.
34. Redfern, M.S., et al., Perceptual inhibition is associated with sensory integration in standing postural control among older adults. The Journals of Gerontology Series B: Psychological Sciences and Social Sciences, 2009: p. gbp060.
35. Nassauer, K.W. and J.M. Halperin, Dissociation of perceptual and motor inhibition processes through the use of novel computerized conflict tasks. Journal of the International Neuropsychological Society, 2003. 9(01): p. 25-30.
36. John, G.H. and P. Langley. Estimating continuous distributions in Bayesian classifiers. in Proceedings of the Eleventh conference on Uncertainty in artificial intelligence. 1995: Morgan Kaufmann Publishers Inc.
37. Quinlan, J.R., C4. 5: programs for machine learning. Vol. 1. 1993: Morgan kaufmann.
38. Quinlan, J.R., Improved use of continuous attributes in C4. 5. arXiv preprint cs/9603103, 1996.
39. Schölkopf, B., et al., New support vector algorithms. Neural computation, 2000. 12(5): p. 1207-1245.
40. Le Cessie, S. and J. Van Houwelingen, Ridge estimators in logistic regression. Applied statistics, 1992: p. 191-201.
41. Grefkes, C., et al., Cortical connectivity after subcortical stroke assessed with functional magnetic resonance imaging. Annals of neurology, 2008. 63(2): p. 236-246.
42. Rehme, A.K., et al., The role of the contralesional motor cortex for motor recovery in the early days after stroke assessed with longitudinal FMRI. Cerebral cortex, 2010: p. bhq140.
43. Wang, L.E., et al., Noradrenergic enhancement improves motor network connectivity in stroke patients. Annals of neurology, 2011. 69(2): p. 375-388.
44. Chen, C.-C., et al., Nonlinear coupling in the human motor system. The Journal of Neuroscience, 2010. 30(25): p. 8393-8399.
45. 林宥辰, 基於虛擬實境復健之中風後運動網路功能性重組研究; Cerebral re-organization of motor networks in response to VR based rehabilitation after stroke. 2014.
46. Lowe, D.G., Distinctive image features from scale-invariant keypoints. International journal of computer vision, 2004. 60(2): p. 91-110.
47. Mikolajczyk, K. and C. Schmid, A performance evaluation of local descriptors. Pattern Analysis and Machine Intelligence, IEEE Transactions on, 2005. 27(10): p. 1615-1630.
48. Bay, H., T. Tuytelaars, and L. Van Gool, Surf: Speeded up robust features, in Computer Vision–ECCV 2006. 2006, Springer. p. 404-417.
49. Hall, M.A. and G. Holmes, Benchmarking attribute selection techniques for discrete class data mining. Knowledge and Data Engineering, IEEE Transactions on, 2003. 15(6): p. 1437-1447.
50. Kohavi, R. and G.H. John, Wrappers for feature subset selection. Artificial intelligence, 1997. 97(1): p. 273-324.
51. Kristeva, R., L. Patino, and W. Omlor, Beta-range cortical motor spectral power and corticomuscular coherence as a mechanism for effective corticospinal interaction during steady-state motor output. Neuroimage, 2007. 36(3): p. 785-792.
52. Chakarov, V., et al., Beta-range EEG-EMG coherence with isometric compensation for increasing modulated low-level forces. Journal of neurophysiology, 2009. 102(2): p. 1115-1120.
53. Finnigan, S. and M.J. van Putten, EEG in ischaemic stroke: Quantitative EEG can uniquely inform (sub-) acute prognoses and clinical management. Clinical neurophysiology, 2013. 124(1): p. 10-19.
54. Diedler, J., et al., Quantitative EEG correlates of low cerebral perfusion in severe stroke. Neurocritical care, 2009. 11(2): p. 210-216.
55. de Vos, C.C., et al., Continuous EEG monitoring during thrombolysis in acute hemispheric stroke patients using the brain symmetry index. Journal of Clinical Neurophysiology, 2008. 25(2): p. 77-82.
56. Brouns, R. and P. De Deyn, The complexity of neurobiological processes in acute ischemic stroke. Clinical neurology and neurosurgery, 2009. 111(6): p. 483-495.
57. Dirnagl, U., C. Iadecola, and M.A. Moskowitz, Pathobiology of ischaemic stroke: an integrated view. Trends in neurosciences, 1999. 22(9): p. 391-397.
58. Hofmeijer, J. and M.J. van Putten, Ischemic Cerebral Damage An Appraisal of Synaptic Failure. Stroke, 2012. 43(2): p. 607-615.
59. Grefkes, C. and G.R. Fink, Reorganization of cerebral networks after stroke: new insights from neuroimaging with connectivity approaches. Brain, 2011: p. awr033.
60. Wang, L., et al., Dynamic functional reorganization of the motor execution network after stroke. Brain, 2010. 133(4): p. 1224-1238.
61. De Vico Fallani, F., et al., Evaluation of the brain network organization from EEG signals: a preliminary evidence in stroke patient. The Anatomical Record, 2009. 292(12): p. 2023-2031.