Spectrotemporal CT data acquisition and reconstruction at low dose.

Area of Science:

• Medical Imaging
• Radiology
• Computational Imaging

Background:

• X-ray computed tomography (CT) is a cornerstone for high-resolution anatomic imaging.
• Expanding CT's utility into functional imaging applications presents significant opportunities.
• Spectral and temporal information can reveal tissue perfusion and dynamic physiological processes.

Purpose of the Study:

• To propose and demonstrate a projection acquisition and reconstruction strategy for 5D CT (3D+dual energy+time).
• To recover spectral and temporal information without substantially increasing radiation dose or sampling time compared to standard anatomic imaging.
• To enable advanced functional imaging applications using CT.

Main Methods:

• A low-rank and sparse matrix decomposition framework was employed for 5D CT reconstruction.
• An explicit rank-sparse signal model was established for spectral and temporal dimensions.
• Rank-sparse kernel regression was used to transfer image structure, enabling approximate separation of temporal and spectral dimensions.
• The split Bregman method and GPU-based implementations were utilized for reconstruction.

Main Results:

• Quantitative 5D simulations and in vivo mouse studies demonstrated successful reconstruction of 20 data sets (10 cardiac phases, 2 energies) with 88 μm isotropic voxels.
• Reconstruction quality was sufficient for material decomposition (gold and iodine maps) to assess myocardial injury and measure cardiac function.
• The 5D CT protocol achieved a 95% reduction in radiation dose and a 40-fold decrease in projection sampling time per cardiac phase and energy.

Conclusions:

• The developed 5D CT acquisition and reconstruction protocol efficiently leverages the rank-sparse nature of spectral and temporal data.
• High-fidelity reconstruction results were achieved without compromising radiation dose or sampling time.
• This method enhances the potential of CT for advanced functional imaging applications.