Related Experiment Video
Updated: Aug 16, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.9K
A clinical and time savings evaluation of a deep learning automatic contouring algorithm.
John S Ginn1, Hiram A Gay1, Jessica Hilliard1
1Department of Radiation Oncology, Washington University School of Medicine, St. Louis, MO 63110, USA.
Summary
Deep learning auto-contouring significantly reduces organ-at-risk (OAR) delineation time in radiation therapy planning. While promising, human review is essential for clinical application.
Area of Science:
- Medical Physics
- Radiotherapy
- Artificial Intelligence
Background:
- Manual organ-at-risk (OAR) contouring is time-consuming in radiation therapy planning.
- Automatic contouring algorithms aim to improve efficiency and consistency.
Purpose of the Study:
- To quantitatively and qualitatively evaluate a deep learning auto-contouring algorithm.
- To assess the time savings associated with the algorithm in clinical workflows.
Main Methods:
- Retrospective analysis of 100 patient cases for quantitative accuracy (Dice, Jaccard, Hausdorff distance).
- Time-motion study comparing manual vs. edited auto-contours by a certified dosimetrist.
- Qualitative evaluation by a radiation oncologist on a 1-4 scoring scale.
Main Results:
- High accuracy (Dice > 0.8) for large organs; low Hausdorff distance for smaller structures.
- Editing auto-contours saved an average of 43.4% (11.8 minutes) per patient.
- Physician review showed >95% of structures scored 3 or 4, indicating good performance.
Conclusions:
- The deep learning auto-contouring algorithm shows potential for reducing clinical contouring time.
- Algorithm performance is promising but requires human oversight and editing before clinical use.
