Generative AI models: the next anaesthetic agent?

Abstract

A study by MacKay and colleagues addresses a pressing need in cardiac anaesthesia by demonstrating an innovative method to extract structured data from free-text intraoperative transoesophageal echocardiography reports. Narrative descriptions of echocardiographic findings are often unstructured, making manual extraction labour-intensive and susceptible to error. By deploying an ensemble of large language models in a consensus-based framework, the authors show that key echocardiographic parameters can be extracted with a high degree of accuracy and manageable error rates. This work presents a technical solution to a specific data-handling challenge and points towards broader applications of artificial intelligence (AI) in streamlining perioperative care.