Combining prior day contours to improve automated prostate segmentation.

Area of Science:

• Radiation Oncology
• Medical Imaging
• Image-guided Radiation Therapy (IGRT)

Background:

• Accurate segmentation of prostate, bladder, and rectum is essential for online adaptive radiotherapy.
• Current methods rely on manual contouring or deformation of previous contours, which can be time-consuming and less accurate.

Purpose of the Study:

• To enhance the accuracy of automated segmentation for prostate, rectum, and bladder contours.
• To improve contouring accuracy on image-guided radiation therapy (IGRT) scans using the Simultaneous Truth and Performance Level Estimation (STAPLE) algorithm by integrating contours from prior IGRT scans.

Main Methods:

• Physician-drawn contours from planning CT and daily kilo-voltage (kV) cone-beam CT (CBCT) scans of six IGRT prostate patients were used.
• Three automated contouring methods were evaluated: Plan (deforming plan CT contours), Previous (deforming previous day's CBCT contours), and STAPLE (combining multiple physician-drawn contours using the STAPLE algorithm).

Main Results:

• STAPLE significantly improved contour accuracy (Dice's coefficient) for bladder, prostate, and rectum compared to Plan and Previous methods.
• STAPLE results demonstrated consistency with intraobserver variability and achieved geometric and dosimetric accuracy suitable for adaptive therapy.
• Automated contouring using STAPLE was time-efficient, with processing times compatible with online adaptive radiotherapy workflows.

Conclusions:

• Combining contours from planning CT and prior CBCT scans via STAPLE is superior to deforming only plan contours for daily CBCT.
• STAPLE enhances contour precision and reduces standard deviation, which is critical for adaptive therapy.
• The STAPLE algorithm enables the automatic generation of geometrically and dosimetrically accurate contours for prostate region kV CBCT within a clinically relevant timeframe.