Multiparametric MRI radiomics and machine learning for outcome prediction in acute ischemic stroke: a retrospective study of 353 patients

 Predictions DON'T GET SURVIVORS RECOVERED! Are you that blitheringly stupid?

Multiparametric MRI radiomics and machine learning for outcome prediction in acute ischemic stroke: a retrospective study of 353 patients

• X

  Xiaoxing Peng ¹

• J

  Jiannan Guo ²

• Q

  Qiuyan Zhou ³

• S

  Shunyu Wang ⁴

• 

  Tingting Li ^5*

• 1. Department of Medical Imaging, Affiliated Hospital of Changchun University of Chinese Medicine, Changchun, Jilin, China

• 2. Department of Electro-diagnosis, Affiliated Hospital of Changchun University of Chinese Medicine, Changchun, Jilin, China

Abstract

Background: 

Acute ischemic stroke (AIS) remains a leading cause of death and disability worldwide. Early prediction for clinical outcomes is essential for guiding therapy.(Wrong! An EXACT DAMAGE DIAGNOSIS is required for mapping to EXACT REHAB POROTOCOLS! All this still needs to be created.)  Multiparametric Magnetic Resonance Imaging (MRI) offers detailed characterization of ischemic tissue and may predict prognosis in AIS patients. This study assesses the prognostic value of multiparametric MRI parameters in predicting functional outcome, recurrence risk and mortality in AIS patients.

Methods: 

We conducted a retrospective analysis of 353 AIS patients who underwent multiparametric MRI within 24 h of onset. MRI metrics including diffusion-weighted imaging (DWI) lesion volume, ADC minimum, perfusion parameters and Susceptibility-Weighted Imaging (SWI) microbleeds were correlated with National Institutes of Health Stroke Scale (NIHSS) and modified Rankin Scale (mRS) scores at discharge and 90 days. Predictive performance for clinical outcomes was assessed by receiver-operating characteristic (ROC) analysis.

Results: 

Among 353 AIS patients with a mean door-to-imaging time of 27 ± 10 min, higher NIHSS, larger DWI volume, prolonged mean transit time (MTT) and Tmax, and increased microbleeds significantly correlated with mortality and disability, while lower ADC, CBF, and CBV correlated inversely with clinical outcomes. NIHSS ≥10.5 and DWI lesion volume ≥52.5 mL yielded AUCs of 0.963 and 0.975 and accuracies of 91.35% and 92.79%, respectively; CBF ≤ 20.5 yielded an AUC of 0.975. NIHSS ≥8.5 and MTT ≥8.5 s showed AUCs of 0.948 and 0.931, respectively, for predicting hemorrhagic transformation. Perfusion thresholds modestly predicted 90-day recurrence (AUC = 0.7). Hypertension, diabetes, and prior stroke were significant clinical predictors of poor outcome. These predictive metrics reflect internal, within-cohort validation only; independent external validation is required before clinical application.

Conclusion: 

Multiparametric MRI provides robust prognostic information in AIS. Quantitative thresholds for perfusion and diffusion parameters help in risk stratification for hemorrhage and mortality, enhancing individualized management and outcome prediction.