Systematic review and meta-analysis of AI for post-stroke epilepsy prediction and diagnosis

This systematic review and meta-analysis examined the utility of artificial intelligence for predicting and diagnosing post-stroke epilepsy. The analysis pooled data from five studies to assess diagnostic performance metrics. No specific population details or setting were reported in the source document.

The primary outcome measured the ability of AI models to identify post-stroke epilepsy. Sensitivity was reported as 88% with a 95% confidence interval of 0.78-0.94. Specificity was reported as 83% with a 95% confidence interval of 0.79-0.86. The area under the summary receiver operating characteristic curve was 0.90 with a 95% confidence interval of 0.87-0.92.

The review did not report absolute numbers, adverse events, or discontinuations. Limitations regarding the certainty of evidence or funding conflicts were not reported. The authors did not provide specific practice relevance recommendations. Clinicians should interpret these pooled metrics with caution given the lack of reported safety data and absolute numbers.