Near-isotropic super-resolution CBCT imaging with a dual-layer flat panel detector.

Area of Science:

• Medical Imaging
• Computerized Tomography
• Deep Learning

Background:

• High spatial resolution is critical for medical imaging.
• Dual-layer flat panel detectors (FPDs) offer enhanced spatial information over single-layer FPDs.
• This enables potential for super-resolution cone beam CT (CBCT) imaging.

Purpose of the Study:

• To investigate the feasibility of achieving near-isotropic super-resolution CBCT imaging using a dual-layer FPD.
• To develop and evaluate a deep learning method for this purpose.

Main Methods:

• Established a mathematical signal model accounting for detector layer shift (Δu, Δv) and gap (Δd).
• Employed a recurrent neural network-based deep neural network (2D-suRi-Net) to retrieve super-resolution information.
• Validated the approach using numerical simulations, a pig leg specimen, and an intersecting cylinder phantom.

Main Results:

• A half-pixel shift (Δu = Δv = 0.5δ) is crucial for super-resolution, especially with detector gaps < 3mm.
• The 2D-suRi-Net effectively retrieved higher spatial resolution from lower-resolution projections.
• Reconstructed images showed <10% spatial resolution difference between axial and coronal planes, indicating near-isotropic capability.

Conclusions:

• Demonstrated the feasibility of near-isotropic super-resolution CBCT imaging with dual-layer FPDs.
• The 2D-suRi-Net method shows promise for enhancing CBCT imaging quality.