Automated contrast-to-noise ratio analysis in chest CT: validation of an open-source segmentation approach

Area of Science:

• Radiology
• Medical Imaging
• Quantitative Analysis

Background:

• Accurate contrast-to-noise ratio (CNR) assessment is crucial for chest CT image quality.
• Manual region-of-interest (ROI) measurements are time-consuming and prone to inter-observer variability.
• Open-source frameworks offer potential for automated image analysis.

Purpose of the Study:

• To evaluate the feasibility and accuracy of automated CNR analysis in chest CT using the open-source body and organ analysis (BOA) framework.
• To validate segmentation modifications for reproducible image-quality assessment.
• To compare automated measurements with manual assessments by radiologists.

Main Methods:

• Retrospective analysis of 100 contrast-enhanced chest CTs (CTA and CTPA).
• Automated BOA segmentations were modified using fat subtraction and binary erosion.
• Comparison of automated and manual measurements using statistical testing, Bland-Altman analysis, and ICC.

Main Results:

• Unmodified BOA segmentations showed significantly lower CNRs compared to manual measurements.
• Optimized segmentation (fat subtraction and binary erosion) achieved no significant difference from radiologist measurements.
• Excellent agreement (ICC 0.89-0.93) and minimal bias were observed in an external validation cohort.

Conclusions:

• A minimally modified open-source segmentation framework enables fully automated, reproducible CNR assessment in chest CT.
• The automated approach achieves expert-level agreement and robust external validation.
• This method streamlines image quality assessment, aids protocol optimization, and supports AI integration.