Soft-Tissue Imaging in Low-Dose, C-Arm Cone-Beam CT Using Statistical Image Reconstruction

Area of Science:

• Medical Imaging
• Radiology
• Computational Imaging

Background:

• C-arm cone-beam CT (CBCT) currently offers limited soft-tissue contrast and resolution.
• Existing imaging tasks are primarily high-contrast, restricting CBCT's utility in soft-tissue interventions.
• Statistical iterative reconstruction techniques represent a significant advancement for improving image quality.

Purpose of the Study:

• To adapt and investigate a penalized likelihood (PL) reconstruction approach for C-arm CBCT.
• To evaluate the performance of PL reconstruction in imaging large, low-contrast soft-tissue structures relevant to surgical guidance.
• To compare PL reconstruction with traditional filtered backprojection (FBP) for soft-tissue imaging tasks.

Main Methods:

• Utilized a mobile C-arm for CBCT acquisition with phantoms and cadavers.
• Employed 3D filtered backprojection (FBP), quadratic PL, and non-quadratic PL reconstruction methods.
• Quantified contrast-to-noise ratio (CNR) and spatial resolution tradeoffs using tissue-equivalent phantoms and anthropomorphic models.

Main Results:

• Statistical reconstruction, particularly non-quadratic PL variants, substantially improved soft-tissue image quality.
• Observed a 2-4 fold increase in CNR at equivalent spatial resolution, with reduced noise and artifacts.
• Demonstrated that non-quadratic penalties and strong regularization could further enhance imaging of large, low-contrast tissues, albeit with some spatial resolution trade-off.

Conclusions:

• Statistical reconstruction techniques offer significant promise for enhancing C-arm CBCT soft-tissue imaging capabilities.
• These methods can overcome conventional tradeoffs between noise, resolution, and radiation dose.
• The findings suggest potential for expanded CBCT applications in soft-tissue interventions, including neurosurgery, thoracic, and abdominal procedures.