Predictive value of cerebral perfusion and guanine nucleotide-binding protein, alpha-stimulating activity polypeptide in ischemic white matter lesions: a machine learning approach.

PURPOSE: To assess the predictive value of guanine nucleotide-binding protein, alpha-stimulating activity polypeptide (GNAS) and cerebral perfusion in various vascular regions for the severity of ischemic white matter lesions (WMLs).

METHODS

Patients hospitalized at the Neurology Department of the Affiliated Hospital of Chengde Medical University between April and November 2023 were evaluated for ischemic cerebral WMLs using magnetic resonance imaging. In this retrospective cohort study, patients were classified into two groups: mild and severe, based on Fazekas scores. White matter perfusion was assessed using image segmentation of arterial spin labeling sequence images. Predictive variables were identified via machine learning (ML). GNAS levels in peripheral blood were measured to explore their association with WML severity.

RESULTS

Among 85 patients (43 mild [24 males and 19 females], 42 severe [27 males and 15 females]), significant differences were observed in age (64.00 ± 8.47 years vs. 68.38 ± 10.85 years, p = 0.041), cerebral atrophy (37.2% vs. 71.4%, p = 0.002), and history of hypertension (41.7% vs. 77.0%, p = 0.002). Corpus callosum perfusion was lower in the severe group (35.84 ± 6.34 vs. 31.73 ± 8.60 mL/[min·100 g], p = 0.037). ML yielded 77.27% model accuracy. Although no significant difference in GNAS levels was observed (p = 0.375), a significant difference was noted in the Fazekas scores (p < 0.001).

CONCLUSION

In patients with ischemic WMLs, factors such as age, sex, history of cerebral infarction, GNAS levels, and specific perfusion metrics are predictive of WML progression. Advanced imaging and ML improve detection. GNAS levels correlated with Fazekas scores, indicating their downregulation in the hypoperfused white matter.