Multi-phase deep learning model for automated disease classification from cardiac cine MRI.

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence

Background:

• Cardiovascular diseases (CVDs) are a leading global cause of mortality.
• Cine Magnetic Resonance Imaging (MRI) is a non-invasive, gold-standard imaging technique for assessing cardiac function and structure.
• Current CVD diagnosis relies on expert interpretation of complex imaging data.

Purpose of the Study:

• To develop and evaluate a deep learning model for automated classification of CVDs using cine MRI.
• To incorporate cardiac function by utilizing both end-diastolic (ED) and end-systolic (ES) phases for improved classification.
• To provide model explainability for enhanced clinical trust and understanding.

Main Methods:

• Investigated single-phase (ED or ES) and multi-phase (ED and ES) deep learning models (ResNet, DenseNet, VGG).
• Trained and tested models on cine MRI datasets for CVD classification.
• Employed explainability techniques to visualize disease-specific regions within the heart.

Main Results:

• Single-phase models achieved test F1-scores of 71.0% (ED) and [Formula: see text] (ES).
• The multi-phase model, incorporating cardiac function, achieved a higher test F1-score of [Formula: see text].
• Explainability visualizations highlighted regions indicative of specific CVDs, aiding interpretation.

Conclusions:

• Deep learning models can effectively automate CVD classification from cine MRI.
• Multi-phase analysis incorporating cardiac function improves classification performance.
• Model explainability is crucial for clinical adoption and trust in AI-driven diagnostics.