AcquisitionFocus: Joint Optimization of Acquisition Orientation and Cardiac Volume Reconstruction Using Deep Learning

Area of Science:

• Medical Imaging
• Artificial Intelligence in Medicine
• Cardiovascular Imaging

Background:

• Cardiac cine MRI is crucial for assessing heart function but is limited by motion artifacts and long acquisition times.
• Achieving high-resolution, volumetric, isotropic data with high temporal resolution in cardiac MRI is constrained by physics.
• Current clinical protocols may not be optimal for comprehensive whole-heart shape assessment within minimal acquisition time.

Purpose of the Study:

• To develop a deep learning model for reconstructing the volumetric shape of cardiac chambers from limited MRI slices.
• To simultaneously optimize slice acquisition orientation for improved shape reconstruction.
• To evaluate the model's performance against standard clinical views in simulated and real cardiac MRI data.

Main Methods:

• A deep learning model was designed to reconstruct 3D cardiac chamber shapes from a reduced set of input MRI slices.
• Slice acquisition orientations were jointly optimized with the shape reconstruction task.
• The model's reconstruction accuracy was compared between standard clinical views and optimized views using metrics like HD95 and Dice scores.

Main Results:

• The proposed deep learning model achieved accurate high-resolution multi-chamber shape reconstruction.
• Reconstruction errors were below 13 mm HD95, and Dice scores exceeded 80%.
• The optimized views demonstrated superior shape reconstruction quality compared to standard clinical views.

Conclusions:

• The deep learning approach effectively reconstructs volumetric cardiac shapes from limited data, addressing motion artifacts in cardiac cine MRI.
• Simultaneous optimization of slice acquisition orientation enhances reconstruction accuracy and efficiency.
• This method shows significant potential for improving diagnostic capabilities in cardiac MRI, especially in cases with diverse pathological shape variations.