動態功能網路應⽤於急性缺⾎型腦中⾵後的運動功能變化之預測;Dynamic functional network to investigate the function change of ischemic stroke patients during rehabilitation

NCUIR > College of Electrical Engineering & Computer Science > Graduate Institute of Electrical Engineering > Electronic Thesis & Dissertation > Item 987654321/93161

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/93161

Title:	動態功能網路應⽤於急性缺⾎型腦中⾵後的運動功能變化之預測;Dynamic functional network to investigate the function change of ischemic stroke patients during rehabilitation
Authors:	吳翌楷;Wu, Yi-Kai
Contributors:	電機工程學系
Keywords:	缺血性腦中風;靜息態功能性磁振造影;動態因果模型;有效性連結;ischemic stroke;resting state functional MRI;dynamic causal modeling;effective connectivity
Date:	2023-01-17
Issue Date:	2024-09-19 16:44:59 (UTC+8)
Publisher:	國立中央大學
Abstract:	本研究提出利用缺血性腦中風個案靜息態磁振造影影像資訊和運動功能量表，建立以腦區間的有效性連結來預測運動功能復原結果的線性回歸模型。建模的步驟包括利用靜息態功能性磁振造影建立運動腦區的動態因果模型、計算腦區之間的有效性連結、計算有效性連結與復健量表之間的相關性、選取與復健量表有顯著相關的有效性連結作為預測模型的特徵、使用所選取的特徵利用MATLAB Regression Learner App加以訓練以建立預測模型。本研究將參與的病患依照中風梗塞位置分群為大腦梗塞、右大腦梗塞、左大腦梗塞、和腦幹梗塞等四個群，然後建立各群專屬的預測模型。這些研究資料是分三期收集，分別是中風剛發作時、中風發作後一個月、和中風發作後三個月;因此預測的時序包含以第一期的有效性連結去預測第二期或第三期的功能量表、以及以第二期的有效性連結去預測第三期的功能量表。研究結果顯示全部預測模型的平均R2達到0.44，未加入臨床變數的預測模型平均R2為0.43，加入年齡、性別的預測模型平均R2為0.48，加入年齡、性別與梗塞體積的預測模型平均R2為0.42。這些結果顯示性別、年齡、梗塞體積等臨床變數加入與否對預測的能力並沒有顯著的影響。此外，四個病患群的平均R2為0.30、0.34、0.61、和0.52。本研究所建立的全部模型中，有一些具有顯著的預測效果，未來也許可以應用於醫師在臨床中風診療、預後、以及照護上的決策之參考或輔助。 ;In this study, we use resting-state functional magnetic resonance imaging (rs-fMRI) of ischemic stroke patients and their motor functional scales to establish linear regression models that predict functional recovery with the effective connectivity between brain regions. The procedure of creating the predictive models including establishing dynamic causal models of motor brain regions with rs-fMRI, computing the effective connectivity between brain regions, evaluating the correlations between the effective connectivity of the edges (i.e., the pairs of brain regions) and the functional scales, selecting the edges corresponding to significant correlations, building predictive models featuring the selected edges, training the models with MATLAB Regression Learner App. According to the infarct location, the patients were divided into four groups: supratentorial, right-hemispheric supratentorial, left-hemispheric supratentorial, and brainstem groups. Predictive models are developed for each group independently. The rs-fMRI were collected at three stages: the time of onset, one month after onset, and three months after onset. Meaningful predictions include predicting stage-2 or stage-3 motor recovery with stage-1 connectivity and predicting stage-3 motor recovery with stage-1 connectivity. The research results showed that the mean R2 was 0.43 for the models consisting of effective connectivity only, 0.48 for the models consisting of effective connectivity, age, and sex, and 0.42 for the models consisting of effective connectivity, age, sex, and infarct volume. The mean R2 of all the models was 0.44. These results suggested that the predictivity was not improved by incorporating clinical parameters such as age and sex in the predictive model. The mean R2 of the four patient groups were 0.30, 0.34, 0.61, and 0.52, respectively. Some of the predictive models established in this research attained statistical significance. These models might be useful in clinical applications to enhance or help post-stroke prognosis and patient care.
Appears in Collections:	[Graduate Institute of Electrical Engineering] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	18	View/Open

社群 sharing

Loading...