Prediction of Stroke Diagnosis Through a Classification Model Based on Cerebral Autoregulation: A Preliminary Study

Ischemic stroke can severely impact the brain perfusion. Cerebral autoregulation (CA) maintains a constant cerebral blood flow during changes in arterial blood pressure. Previous studies demonstrated that CA are compromised with increasing occurrence of stroke. Considering this, this paper aims to evaluate the feasibility of employing
$K$
-nearest neighbor techniqueto automatically predict stroke outcomes, based on CA evaluation by Transfer Function analysis method. One-way ANOVA was performed to verify differences between means of the groups (stroke levels and control). The results presented in this study add to existing evidence that cerebral hemodynamics are negatively affected in stroke patients. The classification algorithm showed promising results, particularly using Transfer Function Analysis with gain (Acc. 68% for very low frequency and Acc 62% for low frequency) and phase parameters (Acc. 62% for very low frequency). But most metrics presented AUC values close to 50% (mean of
$50.7\pm 7.3\%)$
, which indicates that the results have a large percentage of false outcomes. Although further analyzes require improvements in the algorithm, this study showed a promising path in the development of more accurate and reliable diagnostic tools for stroke patients, which can lead to better clinical outcomes.