EchoVisuALL: From Echocardiography to Gene Discovery

Area of Science:

• Cardiovascular research
• Genetics
• Artificial intelligence in medicine

Background:

• Cardiovascular diseases pose a significant global health challenge.
• Advanced phenotyping frameworks are needed for complex mouse genetics studies.
• Current methods for cardiac phenotyping in mice can be labor-intensive and require significant operator expertise.

Purpose of the Study:

• To introduce EchoVisuALL, an AI-enabled pipeline for automated, high-throughput transthoracic echocardiography (TTE).
• To quantify cardiac morphology and function in mice with high reliability and minimal operator dependency.
• To enable the discovery of novel genes associated with cardiac abnormalities.

Main Methods:

• Development of an AI pipeline coupling deep-learning-based left-ventricular segmentation with data reporting.
• Analysis of over 65,000 TTE recordings from more than 18,000 mice, including International Mouse Phenotyping Consortium knockout models.
• Validation against an expert-curated gold standard dataset.
• Extraction of quantitative cardiac parameters across the cardiac cycle and application of multi-dimensional clustering.

Main Results:

• EchoVisuALL quantified cardiac morphology and function with high reliability and minimal operator dependency.
• The pipeline identified 37 out of 715 genes associated with significant cardiac abnormalities.
• These included known human disease genes and 12 previously unrecognized candidate genes (e.g., Cep70, Acot12, Atp8b3).
• Discovered genotype-phenotype associations relate to myocardial energetics, membrane biology, and cardiac remodeling.

Conclusions:

• EchoVisuALL provides a standardized, quantitative foundation for scalable downstream analyses in cardiovascular research.
• The AI pipeline facilitates the discovery of novel cardiac disease genes.
• This approach is crucial for addressing the scale and complexity of contemporary mouse genetics in understanding cardiovascular diseases.