Artificial intelligence for the assessment of diastolic function

Assessment of left ventricular diastolic function remains one of the most challenging aspects of echocardiography. Artificial intelligence (AI) has emerged as a transformative tool capable of automating data acquisition, analysis, and interpretation. This review summarizes recent advances in the use of AI to facilitate diastolic function assessment.

An increasing number of studies have shown the potential for AI-based models to equal or exceed expert-guideline approaches for evaluating diastolic function while improving reproducibility and workflow. Recent trends include the use of more deep learning techniques, reliance on fewer input variables, validation with relevant clinical outcomes, and shift in diastolic function classification from a categorical grading system to a more continuous probabilistic score.

Current guideline-based approaches integrate multiple Doppler, structural, and hemodynamic variables to classify diastolic function. Although these algorithms have improved standardization, they remain limited by interobserver variability, discordant parameters, indeterminate classifications, incomplete datasets, and reduced applicability in complex clinical settings. Machine learning and deep learning approaches can integrate multidimensional echocardiographic features, electrocardiographic signals, and clinical variables to identify latent physiologic patterns beyond conventional rule-based algorithms. Future work will focus on addressing limitations of AI including explainability, generalizability, regulatory considerations, and integration into clinical workflows.