A comparative study based on image quality and clinical task performance for CT reconstruction algorithms in...

Area of Science:

• Medical Imaging
• Radiotherapy Physics
• Computational Imaging

Background:

• Current CT image reconstruction evaluation focuses on dose reduction (ALARA) and image quality metrics (noise, contrast), which are insufficient for clinical radiotherapy task performance.
• Assessing CT reconstruction algorithms requires a method that links image quality, radiation dose, and patient size to radiotherapy task performance.

Purpose of the Study:

• To develop and implement a comparative analysis method for CT reconstruction algorithms, associating image quality, radiation dose, and patient size with radiotherapy task performance.
• To guide the clinical radiotherapy usage of CT reconstruction algorithms, specifically comparing iDose4 iterative reconstruction with filtered back-projection (FBP).

Main Methods:

• Utilized an anthropomorphic pelvis phantom (38-58 cm) and CT images from 34 patients.
• Reconstructed images using FBP (baseline) and iDose4 with varying noise reduction levels.
• Assessed image quality (noise, CNR) and clinical tasks (target CT number, contouring accuracy); conducted qualitative evaluation by radiation oncologists.

Main Results:

• iDose4 achieved up to 66.1% noise reduction and 53.2% CNR improvement compared to FBP.
• iDose4 reduced radiation dose by up to 32% for similar contouring accuracy, with greater improvement in larger patients.
• Radiation oncologists preferred iDose4 images (score 1.21 vs. 3.15 for FBP), recommending moderate noise reduction levels (3-4) for clinical tasks.

Conclusions:

• The iDose4 algorithm offers superior radiation dose control and task performance compared to FBP.
• Moderate noise reduction levels (3-4) in iDose4 are optimal, balancing image quality and task performance, avoiding pixelization and artifacts.
• Quantitative task-based metrics require further investigation for broader clinical applications of CT reconstruction algorithms.