Safety and efficiency of a fully automatic workflow for auto-segmentation in radiotherapy using three commercially...

Area of Science:

• Medical Physics
• Radiotherapy Technology
• Artificial Intelligence in Medicine

Background:

• Radiotherapy auto-segmentation is crucial for efficient treatment planning.
• Current manual workflows can be time-consuming and prone to errors.
• Advancements in deep learning offer potential for automated solutions.

Purpose of the Study:

• To develop and evaluate a standardized, fully automatic workflow for radiotherapy auto-segmentation.
• To compare the safety and efficiency of the automatic workflow against a manual approach.
• To assess the impact of automation on critical failure modes and workflow efficiency.

Main Methods:

• A standardized, fully automatic workflow was developed using three commercial deep learning auto-segmentation applications.
• Safety was evaluated using Failure Mode and Effects Analysis (FMEA).
• Efficiency was measured by the number of mouse clicks required.

Main Results:

• The automatic workflow demonstrated a reduction in eight failure modes, including seven with severity factors ≥7.
• Two failure modes with a Risk Priority Number >125 were also reduced.
• The automatic workflow required zero mouse clicks, indicating significant efficiency gains.

Conclusions:

• The developed automated radiotherapy auto-segmentation workflow significantly improves both safety and efficiency.
• This automation addresses critical failure modes and streamlines the segmentation process.
• The findings support the adoption of automated workflows in clinical radiotherapy practice.