ZeCardioAI: combining zebrafish, AI, and xAI for an in-depth cardiac phenotyping platform

Area of Science:

• Cardiovascular Research
• Computational Biology
• Artificial Intelligence in Medicine

Background:

• Traditional cardiac phenotyping methods face limitations in scalability and objectivity.
• Zebrafish models offer translational relevance and high-throughput potential for cardiovascular research.
• Integrating AI with high-throughput assays can accelerate biological discovery.

Purpose of the Study:

• To develop an AI-powered computational platform (ZeCardioAI) for automated cardiac phenotyping in zebrafish.
• To leverage deep learning (DL) and explainable AI (xAI) for precise and interpretable analysis of cardiac function.
• To apply ZeCardioAI to identify and characterize phenotypic differences in zebrafish models of cardiomyopathy (CM).

Main Methods:

• Development of ZeCardioAI, a computational platform integrating DL and xAI methodologies.
• Automated extraction of comprehensive cardiac phenotypes from live zebrafish imaging.
• Application of machine learning classifiers and xAI for disease classification and feature identification in CM models.

Main Results:

• ZeCardioAI achieved high precision in automatically extracting cardiac phenotypes.
• The platform detected subtle, clinically relevant phenotypic differences in zebrafish models of dilated and hypertrophic cardiomyopathy.
• xAI revealed discriminative features that align with established clinical markers for CM.

Conclusions:

• ZeCardioAI establishes a new standard for AI-powered biological research, enabling large-scale, automated phenotyping.
• The platform has transformative potential for accelerating drug discovery and advancing precision medicine in cardiovascular diseases.
• ZeCardioAI's modular design supports adaptation to diverse disease contexts beyond CMs.