Uncertainty quantification for CT dosimetry based on 10 281 subjects using automatic image segmentation and fast...

Area of Science:

• Medical Physics
• Radiology
• Computational Biology

Background:

• Computed tomography (CT) scans represent a significant source of medical radiation exposure globally.
• Rapidly increasing CT scan frequency, particularly in China, generates vast organ dose data.
• Modern computational methods enable processing large datasets for CT dose uncertainty analysis.

Purpose of the Study:

• To develop and apply a novel method combining automatic image segmentation and GPU-accelerated Monte Carlo (MC) simulations.
• To reconstruct patient-specific organ doses for a large cohort of 10,281 individuals undergoing CT examinations.
• To analyze organ dose distribution patterns and investigate uncertainties in CT dosimetry methods using simplified models.

Main Methods:

• Collected and anonymized CT images and patient health metrics (age, sex, height, weight).
• Utilized deep learning (DeepContour) for automatic organ segmentation.
• Performed GPU-accelerated MC organ dose calculations using GE scanner model and ARCEHR-CT software.
• Conducted statistical analysis on doses for eight key organs.

Main Results:

• Processed data for 10,281 individuals in 16 days using a single GPU.
• Demonstrated profound inter-individual variability in organ doses, even among subjects with similar BMI or WED.
• Identified potential relative errors exceeding 50% in data-fitting methods for CT dosimetry.
• Correlated organ doses with patient metrics (weight, BMI, WED, SSDE), suggesting their use as surrogates with understood uncertainty.
• Observed that CT tube current modulation reduced average doses but did not significantly alter individual organ dose variability.

Conclusions:

• Big data analysis and computational tools facilitate CT dose data mining and uncertainty quantification.
• Significant inter-individual variability in CT organ doses is demonstrable and quantifiable.
• Accounting for inter-individual variability can enhance the accuracy of CT dosimetry.