Deep Learning-Based Cardiac MRI Planning from Localizers to Cine Views Using Landmark Detection.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Cardiology

Background:

• Cardiac MRI planning is crucial for accurate diagnosis.
• Manual planning is time-consuming and prone to variability.
• Deep learning offers potential for automating and improving MRI planning.

Purpose of the Study:

• To evaluate a fully automated deep learning framework for cardiac MRI planning.
• To enhance the efficiency and accuracy of cardiac MRI planning.
• To assess the performance of deep learning models in landmark localization and plane angulation.

Main Methods:

• Retrospective analysis of 1023 cardiac MRI datasets.
• Development of deep learning models trained on expert-annotated landmarks.
• 5-fold cross-validation for model assessment.
• Evaluation using median landmark distances and plane angle differences.

Main Results:

• The deep learning model demonstrated robust performance across all cardiac MRI planes.
• Median landmark localization errors ranged from 4.6 mm to 7.5 mm.
• Angular deviations for short-axis planning were as low as 1.5°.
• Angulation errors for long-axis views were improved using mid-slice SAX compared to base-slice SAX.

Conclusions:

• A deep learning-based automated workflow for cardiac MRI planning is feasible.
• The framework offers improved precision and efficiency.
• This approach has the potential to streamline cardiac MRI procedures.