Related Experiment Video
Updated: Nov 23, 2025
Electroencephalography Network Indices as Biomarkers of Upper Limb Impairment in Chronic Stroke
Published on: July 14, 2023
Structural and functional connectivity of motor circuits after perinatal stroke: A machine learning study
Helen L Carlson1, Brandon T Craig1, Alicia J Hilderley1
1Department of Pediatrics, Cumming School of Medicine, University of Calgary, Calgary, AB, Canada; Calgary Pediatric Stroke Program, Alberta Children's Hospital, Calgary, AB, Canada; Alberta Children's Hospital Research Institute, Calgary, AB, Canada; Hotchkiss Brain Institute, University of Calgary, Calgary, AB, Canada.
Machine learning identified brain connectivity patterns predicting motor skills in children with perinatal stroke. This helps understand individual adaptability and target personalized therapies for better recovery.
Area of Science:
- Neuroscience
- Developmental Biology
- Medical Imaging
Background:
- Developmental neuroplasticity enables young brains to adapt and recover from injury.
- Individual differences in adaptability after early brain injury are not well understood.
- Perinatal stroke offers a model to study neuroplasticity in a developing brain.
Purpose of the Study:
- To utilize machine learning to identify neuroimaging biomarkers predicting motor function in children with perinatal stroke.
- To explore structural and functional brain connectivity patterns associated with motor outcomes.
- To investigate the predictive power of various neuroimaging and demographic features.
Main Methods:
- Machine learning (RELIEFF and random forest regression) was applied to neuroimaging and demographic data.
- Structural connectivity was assessed using white matter tractography of corticospinal tracts.
- Functional connectivity was quantified using blood oxygen-level dependent (BOLD) signal fluctuations.
- Motor function was evaluated using standardized unimanual and bimanual tests.
Main Results:
- Machine learning accurately predicted unilateral motor outcomes using connectivity biomarkers.
- Bimanual motor function prediction required a larger set of features, including connectivity from both hemispheres.
- Cortical and subcortical regions, along with interhemispheric connectivity, played significant roles.
- Lesion characteristics and age at scan were predictive but less influential than connectivity patterns.
Conclusions:
- Machine learning effectively identifies neuroimaging biomarkers for motor function in perinatal stroke.
- Connectivity patterns across both hemispheres are crucial for motor recovery.
- These findings can guide personalized neuromodulation targets for improved outcomes in affected children.
