Thin-Slice Brain CT Image Quality and Lesion Detection Evaluation in Deep Learning Reconstruction Algorithm.

Area of Science:

• Radiology and Medical Imaging
• Artificial Intelligence in Healthcare
• Neuroimaging Techniques

Background:

• Clinical use of AI-based Precise Image (PI) algorithms in brain imaging is not well-established.
• PI is a deep-learning reconstruction (DLR) technique designed to reduce noise in low-dose CT scans while maintaining a Filtered Back Projection (FBP)-like image appearance.
• This study evaluates PI's efficacy against Iterative Reconstruction (IR) and FBP for thin-slice brain CT.

Purpose of the Study:

• To compare the effectiveness of PI, IR, and FBP in enhancing image quality.
• To assess the ability of these reconstruction methods to improve lesion detection in 1.0 mm thin-slice brain CT images.
• To evaluate PI's performance in maintaining image quality at low radiation doses.

Main Methods:

• Retrospective analysis of 60 non-contrast brain CT scans.
• Reconstruction using four methods: routine 5.0 mm FBP, 1.0 mm thin-slice FBP, 1.0 mm thin-slice IR, and 1.0 mm thin-slice PI.
• Subjective image quality assessment via Likert scale by radiologists and objective metrics (CNR, SNR, noise) were analyzed.

Main Results:

• Thin-slice PI images exhibited the lowest noise and artifacts, with the highest contrast-to-noise ratio (CNR) and signal-to-noise ratio (SNR) (p < 0.001).
• Both PI and IR significantly improved image quality over routine FBP (p < 0.05).
• PI achieved superior lesion conspicuity and diagnostic confidence, with a 100% detection rate for lacunar lesions, outperforming other methods.

Conclusions:

• Precise Image (PI) reconstruction significantly enhances image quality and lesion detectability in thin-slice brain CT.
• PI demonstrates superior performance compared to Iterative Reconstruction (IR) and Filtered Back Projection (FBP).
• PI holds potential as a new clinical standard for brain CT imaging.