Machine Learning Assisted Stroke Prediction in Mechanical Circulatory Support: Predictive Role of Systemic...

Area of Science:

• Cardiology
• Neurology
• Mitochondrial Biology

Background:

• Stroke is a significant risk in advanced congestive heart failure (CHF) patients receiving continuous-flow left ventricular assist devices (CF-LVADs).
• Mitochondrial oxidative phosphorylation (OxPhos) abnormalities are linked to neurodegeneration and cerebral ischemia.

Purpose of the Study:

• To investigate if prior stroke is associated with systemic mitochondrial OxPhos abnormalities in CF-LVAD patients.
• To determine if these abnormalities are more pronounced in patients at risk for new strokes post-implantation.

Main Methods:

• Studied 50 CF-LVAD patients (25 with prior stroke, 25 without).
• Measured OxPhos complex proteins (C.I-C.V) in blood leukocytes pre- and post-CF-LVAD implantation.
• Utilized machine learning (LASSO, Random Forest) for stroke prediction modeling.

Main Results:

• The prior-stroke group exhibited significantly lower levels of OxPhos proteins (C.I, C.II, C.IV, C.V) compared to the no-prior-stroke group, both pre- and post-CF-LVAD.
• OxPhos proteins further decreased in the prior-stroke group post-CF-LVAD.
• Machine learning models identified six prognostic factors, achieving an AUC of 0.93 for predicting postoperative stroke.

Conclusions:

• Mitochondrial dysfunction, indicated by peripheral OxPhos protein reduction, exists in CHF patients with prior ischemic stroke even before CF-LVAD implantation.
• Reduced OxPhos protein expression may serve as a predictive biomarker for new strokes following CF-LVAD implantation.