Accuracy of patient-specific organ dose estimates obtained using an automated image segmentation algorithm

Area of Science:

• Medical Physics
• Radiology
• Computational Imaging

Background:

• Accurate estimation of patient-specific organ doses from computed tomography (CT) scans is crucial for radiation protection.
• Manual segmentation of organs for dose calculation is time-consuming and prone to inter-observer variability.
• Automated methods are needed to improve the efficiency and consistency of organ dose estimation in CT.

Purpose of the Study:

• To develop and validate a rapid, accurate, and automated software tool for estimating patient-specific organ doses from CT scans.
• To quantify the accuracy of organ dose estimates derived from an automated segmentation algorithm compared to expert segmentation.
• To assess the impact of segmentation errors on mean organ dose calculations.

Main Methods:

• Development of an automated software tool integrating simulation-based dose mapping and automated segmentation algorithms.
• A leave-one-out validation study using 20 head-neck CT scans segmented by experts into nine regions.
• Comparison of mean organ doses calculated from automated and expert segmentations using Monte Carlo-generated dose maps.

Main Results:

• The automated segmentation algorithm achieved mean organ dose estimates within 10% of expert segmentation for most regions.
• Median errors for organ dose estimates were below 2% for most regions.
• The spinal canal region showed higher median errors ([Formula: see text]) with maximum absolute errors of 28% (single-atlas) and 11% (multi-atlas).

Conclusions:

• The automated segmentation algorithm demonstrates sufficient accuracy for reliable patient-specific organ dose estimation from CT scans.
• Despite minor segmentation inaccuracies, the automated tool provides clinically relevant organ dose estimates.
• This automated approach offers a rapid and accurate solution for dosimetry in CT, improving radiation safety protocols.