Related Experiment Video
Updated: Jul 15, 2026
08:23
A Multimodal Imaging- and Stimulation-based Method of Evaluating Connectivity-related Brain Excitability in Patients with Epilepsy
Published on: November 13, 2016
Machine learning combined with resting-state functional MRI to characterize functional brain differences in
Yuanxin Shao1,2, Chao Liang2, Dan Xu3
1First Clinical Medical College, Nanjing University of Chinese Medicine, Nanjing, Jiangsu, China.
Frontiers in Psychiatry
|July 13, 2026
Summary
Post-stroke depression (PSD) shows distinct resting-state functional MRI differences in brain regions like the cingulate and thalamus. Machine learning identified key imaging features, aiding in understanding PSD neurobiology.
Area of Science:
- Neuroimaging
- Neuroscience
- Machine Learning in Medicine
Background:
- Post-stroke depression (PSD) is a frequent complication following stroke.
- Resting-state functional imaging correlates of PSD are not fully understood.
- This study investigates multi-level functional brain differences in PSD patients.
Purpose of the Study:
- To identify resting-state functional MRI differences between PSD patients and healthy controls.
- To evaluate interpretable machine learning for identifying PSD-associated imaging features.
- To explore the neurobiological underpinnings of PSD.
Main Methods:
- Resting-state functional MRI was performed on 50 PSD patients and 50 controls.
- Four imaging indices (ALFF, ReHo, DC, FC) were extracted using the AAL atlas.
- LASSO regression and nine machine-learning classifiers were used, with SHAP for feature interpretation.
Main Results:
- PSD patients exhibited widespread resting-state functional differences in cingulate, thalamic, prefrontal, and other regions.
- Twenty-nine features differed between groups; LASSO identified 10 core features (AUC 0.878).
- The Extra Trees model achieved an AUC of 0.889, with key features including left anterior cingulate DC and left thalamus ReHo.
Conclusions:
- PSD is associated with multi-level resting-state functional brain differences.
- Interpretable machine learning successfully identified neurobiologically relevant rs-fMRI features for PSD.
- Further validation in larger cohorts is needed to confirm specificity and clinical utility.
