Prognostic value of multi-PLD ASL radiomics in acute ischemic stroke

Area of Science:

• Neuroimaging
• Machine Learning
• Radiomics

Background:

• Early prediction of acute ischemic stroke (AIS) prognosis is crucial for personalized treatment planning.
• Arterial spin labeling (ASL) provides non-contrast-enhanced perfusion imaging.
• Radiomics extracts quantitative features from medical images, offering potential for predictive modeling.

Purpose of the Study:

• To develop and validate a machine learning (ML) model for early and precise prediction of AIS prognosis.
• To investigate the utility of multi-post-labeling delay (multi-PLD) ASL radiomics features combined with clinical data.

Main Methods:

• 102 AIS patients were enrolled, with clinical data (age, NIHSS) and multi-PLD ASL images collected.
• Radiomics features were extracted from cerebral blood flow (CBF) images, selected using LASSO regression.
• Three models (clinical, CBF radiomics, combined) were built using eight ML algorithms and evaluated with ROC curves and DCA.

Main Results:

• The combined ML model using extreme gradient boosting achieved the highest predictive performance with an AUC of 0.876.
• The combined model significantly outperformed the clinical (AUC=0.658) and radiomics-only (AUC=0.755) models.
• Key predictors included baseline NIHSS score, age, and CBF texture and shape features.

Conclusions:

• Integrating clinical data with multi-PLD ASL radiomics offers a robust method for AIS prognosis prediction.
• This non-contrast approach is valuable for patients with contraindications to contrast agents.
• The developed model supports clinicians in creating individualized treatment strategies to improve patient outcomes.