Artificial Intelligence-Based Autosegmentation: Advantages in Delineation, Absorbed Dose-Distribution, and Logistics.

Area of Science:

• Medical Imaging and Radiation Oncology
• Artificial Intelligence in Healthcare
• Clinical Workflow Optimization

Background:

• Accurate delineation of organs-at-risk (OARs) and clinical target volumes is crucial for effective radiation therapy.
• Manual contouring is time-consuming and subject to inter-observer variability.
• AI offers potential for automating and improving the accuracy of the contouring process.

Purpose of the Study:

• To compare the performance of AI auto-contouring against human practitioners.
• To evaluate precision, dose distribution differences, and time efficiency of AI vs. manual contouring.
• To assess the impact of AI-assisted contouring on radiation therapy planning.

Main Methods:

• Retrospective analysis of 3D CT datasets from 75 patients across head and neck, breast, and prostate cancer segments.
• Comparison of contours generated by an experienced radiation oncologist (MD), AI alone, and AI with manual corrections (AI+C).
• Evaluation using Dice Similarity Coefficient (DSC) for accuracy and assessment of dose-volume parameters and time consumption.

Main Results:

• AI and AI+C achieved mean DSC scores > 0.7 for 74% and 80% of structures, respectively.
• AI-assisted contouring led to significant time savings: 68% (breast), 51% (prostate), and 71% (head and neck).
• Minimal clinically relevant differences in OAR exposure were observed with AI-assisted contouring.

Conclusions:

• AI effectively generates clinically acceptable OARs and target volumes across various anatomical sites.
• The AI+C workflow is efficient, requiring minimal corrections and offering substantial time savings.
• AI in auto-contouring represents a valuable tool for optimizing radiation therapy planning and resource allocation.