A machine learning-based predictive model for predicting early neurological deterioration in lenticulostriate atheromatous disease-related infarction.

BACKGROUND AND AIM

This study aimed to develop a predictive model for early neurological deterioration (END) in branch atheromatous disease (BAD) affecting the lenticulostriate artery (LSA) territory using machine learning. Additionally, it aimed to explore the underlying mechanisms of END occurrence in this context.

METHODS

We conducted a retrospective analysis of consecutive ischemic stroke patients with BAD in the LSA territory admitted to Dongyang People's Hospital from January 1, 2018, to September 30, 2023. Significant predictors were identified using LASSO regression, and nine machine learning algorithms were employed to construct models. The logistic regression model demonstrated superior performance and was selected for further analysis.

RESULTS

A total of 380 patients were included, with 268 in the training set and 112 in the validation set. Logistic regression identified stroke history, systolic pressure, conglomerated beads sign, middle cerebral artery (MCA) shape, and parent artery stenosis as significant predictors of END. The developed nomogram exhibited good discriminative ability and calibration. Additionally, the decision curve analysis indicated the practical clinical utility of the nomogram.

CONCLUSION

The novel nomogram incorporating systolic pressure, stroke history, conglomerated beads sign, parent artery stenosis, and MCA shape provides a practical tool for assessing the risk of early neurological deterioration in BAD affecting the LSA territory. This model enhances clinical decision-making and personalized treatment strategies.