Static and temporal dynamic changes in brain activity in patients with post-stroke balance dysfunction: a pilot resting state fMRI

 I see NOTHING HERE THAT GETS SURVIVORS RECOVERED! Fucking useless! It describes something but does nothing for survivors! I'd have you all fired!

Static and temporal dynamic changes in brain activity in patients with post-stroke balance dysfunction: a pilot resting state fMRI

Zhiqing Tang^1,2Tianhao Liu^1,2Junzi Long^1,2Weijing Ren³Ying Liu^1,2Hui Li^2,4Kaiyue Han^1,2Xingxing Liao^1,2Xiaonian Zhang^1,2Haitao Lu^1,2*Hao Zhang^1,2*

• ¹School of Rehabilitation, Capital Medical University, Beijing, China
• ²Beijing Bo'ai Hospital, China Rehabilitation Research Center, Beijing, China
• ³Department of Rehabilitation, Shandong Provincial Hospital Affiliated to Shandong First Medical University, Jinan, Shandong, China
• ⁴Cheeloo College of Medicine, Shandong University, Jinan, Shandong, China

Objective: The aim of this study was to investigate the characteristics of brain activity changes in patients with post-stroke balance dysfunction and their relationship with clinical assessment, and to construct a classification model based on the extreme Gradient Boosting (XGBoost) algorithm to discriminate between stroke patients and healthy controls (HCs).

Methods: In the current study, twenty-six patients with post-stroke balance dysfunction and twenty-four HCs were examined by resting-state functional magnetic resonance imaging (rs-fMRI). Static amplitude of low frequency fluctuation (sALFF), static fractional ALFF (sfALFF), static regional homogeneity (sReHo), dynamic ALFF (dALFF), dynamic fALFF (dfALFF) and dynamic ReHo (dReHo) values were calculated and compared between the two groups. The values of the imaging metrics for the brain regions with significant differences were used in Pearson correlation analyses with the Berg Balance Scale (BBS) scores and as features in the construction of the XGBoost model.

Results: Compared to HCs, the brain regions with significant functional abnormalities in patients with post-stroke balance dysfunction were mainly involved bilateral insula, right fusiform gyrus, right lingual gyrus, left thalamus, left inferior occipital gyrus, left inferior temporal gyrus, right calcarine fissure and surrounding cortex, left precuneus, right median cingulate and paracingulate gyri, right anterior cingulate and paracingulate gyri, bilateral supplementary motor area, right putamen, and left cerebellar crus II. XGBoost results show that the model constructed based on static imaging features has the best classification prediction performance.

Conclusion: In conclusion, this study provided evidence of functional abnormalities in local brain regions in patients with post-stroke balance dysfunction. The results suggested that the abnormal brain regions were mainly related to visual processing, motor execution, motor coordination, sensorimotor control and cognitive function, which contributed to our understanding of the neuropathological mechanisms of post-stroke balance dysfunction. XGBoost is a promising machine learning method to explore these changes.

1 Introduction

Stroke can cause a variety of neurological impairments, including sensory, cognitive, and motor impairments, poor coordination, and difficulty maintaining balance (Winstein et al., 2016). More than 80% of stroke survivors experience balance dysfunction, which can limit their ability to participate in daily activities and significantly reduce their quality of life (Schmid et al., 2013; Tyson et al., 2006). Balance dysfunction is strongly associated with an increased risk of falls in stroke patients and is also recognized as an important factor affecting patients’ ability to walk independently (Nayak et al., 2024; Park et al., 2021). However, the underlying brain mechanisms of post-stroke balance dysfunction remain unclear (Peng et al., 2024). Therefore, there is a need to clarify the brain function abnormalities in patients with post-stroke balance dysfunction, which may help to develop precise therapeutic interventions.

Resting-state functional magnetic resonance imaging (rs-fMRI), which measures low-frequency fluctuations in blood oxygen level-dependent (BOLD) signals, is a promising tool for studying spontaneous brain activity and has been widely used to study changes in brain function in both patients and healthy individuals (Biswal et al., 1995; Raimondo et al., 2021). A large number of studies have shown that low-frequency fluctuations are critical for understanding human brain activity (Auer, 2008; Lee et al., 2013). Various methods such as amplitude of low frequency fluctuation (ALFF), fractional ALFF (fALFF), and regional homogeneity (ReHo) have been widely used to analyze changes in brain function after stroke (Quan et al., 2022; Wang H. et al., 2022; Wu et al., 2023)]. ALFF reflects the intensity of intrinsic brain activity by measuring spontaneous neural activity in localized areas of the brain in the range of 0.01 ~ 0.1 Hz (Zang et al., 2007). Later, Zou et al. proposed fALFF based on ALFF, i.e., the ratio of the low-frequency power spectrum to the power spectrum of the whole frequency range, which reflects the relative contribution of specific low-frequency oscillations to the whole detectable frequency range (Zou et al., 2008). ReHo, calculated on the basis of Kendall’s coefficients, was used to measure the similarity of a given voxel’s time series to its nearest neighbor and to detect subtle changes in neural activity in specific brain regions (Zang et al., 2004). The combination of the above three methods can more comprehensively reflect the spontaneous neural activity of the brain. Although a large number of studies have detected significant alterations in ALFF, fALFF, and ReHo in some brain regions after stroke (Li et al., 2022; Yang et al., 2024; Zhao et al., 2018), studies exploring alterations in brain function associated with balance dysfunction are still lacking.

Although it is well known that brain activity changes dynamically (Wang et al., 2023), most current studies have traditionally calculated indicators such as ALFF under the assumption that the BOLD signal remains constant throughout the functional magnetic resonance imaging (fMRI) scan, ignoring the fact that local brain activity has dynamic properties during time-varying processes and may therefore miss valuable information (Cohen, 2018; Xie et al., 2018). Previous studies have suggested that dynamic analyses can compensate for the shortcomings of static analyses and that a combination of the two may be more conducive to a more comprehensive understanding of the neuropathological changes in disease (Bonkhoff et al., 2020; Bonkhoff et al., 2021; Wang et al., 2023). The sliding window approach, the main method of dynamic analysis techniques, is considered effective and sensitive in exploring the temporal variability of brain activity and has been widely used to study abnormal brain function in neurological and psychiatric disorders (Cui et al., 2020; Liu et al., 2021). Some studies have found significant changes in dynamic ALFF and other indicators in stroke patients that correlate significantly with clinical characteristics (Chen and Li, 2023; Wang et al., 2023). However, there are few studies using both dynamic and static analysis methods to investigate brain functional activity in patients with post-stroke balance dysfunction.

In addition, machine learning methods are powerful tools for the classification of patients with respect to healthy controls (Ruksakulpiwat et al., 2023; Wang J. et al., 2022). There have been a number of neuroimaging studies applying machine learning methods to detect biomarkers of disease and build classification or prediction models. Extreme Gradient Boosting (XGBoost) is a well-established and widely used machine learning modelling algorithm for solving supervised learning problems using the gradient boosting framework, which is highly accurate, difficult to overfit and scalable (Chen et al., 2023; Hu et al., 2022). As a decision tree-based algorithm, XGBoost was named the best algorithm in the Machine Learning and Prediction Competition hosted by Kaggle.com (Chen and Guestrin, 2016). XGBoost has been gradually applied to the medical field and has demonstrated superior model performance compared to other machine learning algorithms such as logistic regression, support vector machines and random forests in many studies (Ai et al., 2024; Hou et al., 2020; Tang et al., 2024).

In this study, firstly, based on rs-fMRI data, static and dynamic metrics, including static ALFF (sALFF), static fALFF (sfALFF), static ReHo (sReHo), dynamic ALFF (dALFF), dynamic fALFF (dfALFF), and dynamic ReHo (dReHo), were used to investigate the characteristics of brain activity changes in patients with post-stroke balance dysfunction. Secondly, the relationship between imaging metrics and clinical assessment were explored. Finally, the values of the imaging metrics of the brain regions with significant differences were used as features for feature screening and classification model construction using the XGBoost algorithm.

More at link.