dc.description.abstract | 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.
| en_US |