One-mSv CT colonography: Effect of different iterative reconstruction algorithms on radiologists' performance

Area of Science:

• Radiology
• Medical Imaging
• Computational Imaging

Background:

• Ultra-low dose CT colonography (CTC) aims to reduce radiation exposure.
• Standard reconstruction algorithms like filtered back projection (FBP) may limit image quality at low doses.
• Iterative reconstruction (IR) algorithms offer potential for noise reduction and improved image quality.

Purpose of the Study:

• To evaluate the impact of different reconstruction algorithms (FBP, ASIR, Veo) on image noise.
• To assess the effect of these algorithms on radiologists' performance in detecting polyps during ultra-low dose CTC.
• To determine the feasibility of achieving diagnostic quality at approximately one milli-sievert (mSv) radiation dose.

Main Methods:

• Retrospective analysis of 28 patients undergoing CTC and colonoscopy.
• CTC scans acquired at ultra-low dose (fixed 10mAs) and reconstructed using FBP, ASIR, and Veo algorithms.
• Image noise quantified using ANOVA; polyp detection evaluated by two readers using per-polyp sensitivity and JAFROC analysis.

Main Results:

• Veo reconstruction demonstrated significantly lower image noise compared to ASIR and FBP.
• Per-polyp sensitivity and figure-of-merit for polyp detection were highest with Veo, followed by ASIR, and then FBP.
• Veo and ASIR showed statistically significant improvements over FBP for polyp detection by both novice and expert readers.

Conclusions:

• Standard FBP algorithm is insufficient for diagnostic quality at one-mSv CTC.
• Iterative reconstruction algorithms, especially Veo, significantly enhance diagnostic performance (sensitivity and figure-of-merit) in low-dose CTC.
• IR algorithms are crucial for enabling effective ultra-low dose CTC examinations.