Cerebral blood flow alterations and machine learning classification in chronic low back pain using multi-PLD ASL...

Area of Science:

• Neuroimaging
• Radiology
• Machine Learning

Background:

• Chronic low back pain (CLBP) is a widespread condition with poorly understood causes and significant societal impact.
• Previous studies on cerebral blood flow (CBF) alterations in CLBP using single post-labeling delay (PLD) arterial spin labeling (ASL) have produced inconsistent findings.

Purpose of the Study:

• To investigate CBF alterations in CLBP patients using multi-PLD ASL.
• To explore the potential of machine learning in classifying CLBP based on CBF radiomic features.

Main Methods:

• Seventy-eight CLBP patients and 78 healthy controls underwent multi-PLD ASL scans.
• Voxel-wise comparisons of normalized CBF were conducted, followed by correlation analyses with clinical pain and disability measures.
• Machine learning models (XGBoost) were trained using radiomic features from brain regions with significant CBF differences, employing nested cross-validation and LASSO feature selection.

Main Results:

• CLBP patients showed significant hyperperfusion in the right lingual gyrus and right thalamus compared to controls.
• CBF in the right lingual gyrus correlated positively with Oswestry Disability Index scores.
• Thalamic CBF positively correlated with pain intensity, and the XGBoost model achieved an AUC of 0.842 for CLBP classification.

Conclusions:

• Region-specific CBF alterations in the lingual gyrus and thalamus are linked to pain severity and functional impairment in CLBP.
• Machine learning analysis of CBF radiomic features demonstrates promising discriminative ability for identifying CLBP patients.