Enhancing Cone-Beam CT Image Quality in TIPSS Procedures Using AI Denoising

Area of Science:

• Medical Imaging
• Radiology
• Artificial Intelligence in Medicine

Background:

• Transjugular Intrahepatic Portosystemic Shunt (TIPS) procedures require high-quality imaging.
• Radiation dose and motion artifacts are significant challenges in CBCT for TIPS.
• Optimizing the balance between image quality, radiation dose, and scan time is crucial.

Purpose of the Study:

• To evaluate a deep learning-based denoising algorithm for CBCT in TIPS procedures.
• To assess the algorithm's ability to reduce radiation dose and motion artifacts.
• To maintain diagnostic image quality with shorter acquisition times.

Main Methods:

• Retrospective analysis of 44 TIPS patients' CBCT data.
• Comparison of 6-second and 3-second acquisition times.
• Image reconstruction using traditional filtered back projection (Original) and an AI denoising algorithm (AID).
• Objective (CNR, noise) and subjective (radiologist evaluation) image quality assessments.

Main Results:

• Shorter 3-second scans significantly reduced radiation dose (DAP) compared to 6-second scans.
• The AI denoising algorithm (AID) achieved the highest Contrast-to-Noise Ratio (CNR) in 6-second scans.
• No significant difference in noise levels between 6-second Original and 3-second AID images.
• Subjective assessments showed superior quality for 6-second AID images, with no significant difference between 6-second Original and 3-second AID.

Conclusions:

• The AI denoising algorithm effectively enhances CBCT image quality for TIPS.
• Shorter CBCT acquisition times are feasible with AI denoising, reducing radiation exposure.
• AI-driven denoising offers a promising solution for optimizing image quality and safety in TIPS procedures.