AIRPORT: A Data Consistency Constrained Deep Temporal Extrapolation Method To Improve Temporal Resolution In Contrast...

Area of Science:

• Medical Imaging
• Computational Imaging
• Artificial Intelligence in Medicine

Background:

• Traditional tomographic reconstruction necessitates static objects, leading to temporal-averaging errors in dynamic imaging.
• Reconstructing images from narrower angular ranges improves temporal resolution but introduces data-insufficiency errors.

Purpose of the Study:

• To decouple the trade-off between temporal-averaging and data-insufficiency errors in contrast-enhanced CT.
• To enable high-temporal-resolution imaging of dynamic objects without compromising image quality.

Main Methods:

• Developed a data consistency constrained deep temporal extrapolation method (AIRPORT).
• Applied AIRPORT to contrast-enhanced computed tomography (CT) imaging scenarios.
• Utilized single short-scan data acquisition for non-sparse imaging tasks.

Main Results:

• Accurate reconstruction of time-varying objects was achieved.
• Demonstrated a temporal resolution of 40 frames per second.
• AIRPORT effectively mitigates errors in dynamic CT imaging.

Conclusions:

• AIRPORT successfully decouples temporal-averaging and data-insufficiency errors in CT.
• The method enables high-fidelity, high-temporal-resolution dynamic CT imaging.
• AIRPORT is applicable to general non-sparse imaging tasks with single short-scan acquisition.