Artificial intelligence-assisted peri-operative echocardiography: a multicentre observational study.

Area of Science:

• Cardiology
• Anesthesiology
• Medical Imaging
• Artificial Intelligence

Background:

• Point-of-care transthoracic echocardiography (POCUS) aids peri-operative management but faces time and skill constraints.
• Artificial intelligence (AI) offers potential to streamline POCUS assessment for clinicians and patients.

Purpose of the Study:

• To evaluate the agreement between AI-driven echocardiographic analysis and manual measurements by clinicians.
• To assess the feasibility of using AI for key echocardiographic parameter classification in pre-operative settings.

Main Methods:

• A prospective, multicentre observational study enrolled adult patients undergoing pre-operative echocardiography.
• Certified anaesthetists acquired a 12-view echocardiographic protocol, uploaded to the US2.AI platform for automated analysis.
• Manual clinician measurements served as the reference standard for comparison.

Main Results:

• AI analysis of 202 patients showed good to excellent agreement (ICC 0.605-0.956) with clinician measurements for most parameters.
• US2.AI correctly classified left ventricular systolic function (91%) and diastolic dysfunction (96%).
• High correlations were observed for ventricular size/function and atrial size; AI correctly identified all cases of pulmonary hypertension and severe aortic stenosis.

Conclusions:

• AI analysis using a limited echocardiographic sequence provides essential peri-operative information.
• High agreement between US2.AI and clinicians for 10 parameters supports its clinical utility.
• Integration of US2.AI into peri-operative echocardiography workflows is supported, pending outcome studies.