Improving head and neck CTA with hybrid and model-based iterative reconstruction techniques

Area of Science:

• Medical Imaging
• Radiology
• Image Reconstruction

Background:

• Computed tomography angiography (CTA) is crucial for visualizing head and neck vasculature.
• Image reconstruction algorithms significantly impact CTA image quality and diagnostic accuracy.
• Comparing traditional filtered back projection (FBP) with iterative reconstruction methods (hybrid iterative reconstruction [HIR] and model-based iterative reconstruction [MIR]) is essential for optimizing imaging protocols.

Purpose of the Study:

• To compare the objective and subjective image quality of head and neck CTA reconstructed using FBP, HIR, and MIR algorithms.
• To evaluate the impact of different reconstruction algorithms on key imaging metrics like contrast-to-noise ratio (CNR), vascular contrast, automated vessel analysis (AVA), and stenosis grading.

Main Methods:

• Retrospective analysis of 34 head and neck CTA datasets.
• Reconstruction of data using filtered back projection (FBP), hybrid iterative reconstruction (HIR - iDose(4)), and a prototype model-based iterative reconstruction (MIR - IMR) algorithm.
• Objective quantitative analysis (CNR, vascular contrast, AVA, stenosis grade) and subjective qualitative assessment (image quality ranking) were performed.

Main Results:

• MIR algorithms demonstrated significantly higher vascular contrast and improved automated vessel analysis completeness compared to FBP and HIR (p<0.0001).
• The highest contrast-to-noise ratio (CNR) was achieved with high MIR, followed by low MIR, high HIR, mid HIR, and FBP (p<0.001).
• Subjective image quality was highest with high HIR at the circle of Willis and carotid bifurcation, while low MIR and high HIR were preferred at the shoulder level.

Conclusions:

• Model-based iterative reconstruction (MIR) significantly enhances objective image quality in head and neck CTA, reducing noise and improving automated vessel analysis.
• Filtered back projection (FBP) resulted in the lowest objective image quality.
• Hybrid iterative reconstruction (HIR) provided superior subjective image quality in specific anatomical regions, highlighting the trade-offs between objective and subjective assessments.