A machine learning-based pipeline for multi-organ/tissue patient-specific radiation dosimetry in CT

Area of Science:

• Medical Physics
• Radiology
• Computational Biology

Background:

• Personalized radiation dosimetry in CT is crucial for patient safety but is often time-consuming.
• Accurate estimation of organ and tissue radiation doses is essential for optimizing treatment plans and minimizing side effects.

Purpose of the Study:

• To develop a machine learning-based pipeline for multi-organ/tissue personalized radiation dosimetry in CT.
• To create an automated workflow for estimating patient-specific radiation doses in CT examinations.

Main Methods:

• Retrospective collection of 95 chest and 85 abdominal CT scans.
• Development of organ/tissue-specific dose prediction deep neural networks (DNNs) using Monte Carlo (MC) simulations and CT data.
• Integration of an open-source organ segmentation tool with DNNs for automated dose estimation in 30 organs/tissues.

Main Results:

• The pipeline achieved low median percentage differences between MC-derived doses and DNN predictions for various organs (e.g., lung vessels 4.3%, small bowel 4.7%).
• Statistical analysis indicated no significant differences between predicted and actual doses (p > 0.18).
• The mean inference time for the validation cohort was 77.0 ± 11.0 seconds, demonstrating time efficiency.

Conclusions:

• The proposed workflow enables fast and accurate organ/tissue radiation dose estimations in CT.
• The developed algorithms and DNNs are publicly available, facilitating adoption in clinical practice.
• This automated pipeline serves as a valuable tool for patient-specific dosimetry in routine CT procedures.