Automatic tandem and ring reconstruction using MRI for cervical cancer brachytherapy

Area of Science:

• Medical physics
• Oncology
• Radiotherapy

Background:

• MRI-guided brachytherapy for cervical cancer offers superior soft-tissue contrast.
• Conventional metallic fiducials cannot be used for source path reconstruction in MRI-guided brachytherapy.
• Manual applicator reconstruction is time-consuming and hinders automated treatment planning.

Purpose of the Study:

• To develop and validate an algorithm for automatic reconstruction of tandem and ring applicators using MRI.
• To improve the efficiency and automation of cervical cancer brachytherapy treatment planning.

Main Methods:

• A model-to-image registration algorithm was used with T2-weighted MR images.
• The algorithm involved image filtering, maximum intensity projection, and Hough transform for applicator identification.
• Three-dimensional surface model registration was optimized using image intensity gradients.
• Algorithm performance was compared against manual reconstructions by two independent observers.

Main Results:

• The algorithm demonstrated variability and accuracy within the range of interobserver variability.
• Mean algorithm variability for the ring and tandem applicators was 0.72 ± 0.32 mm and 0.70 ± 0.29 mm, respectively.
• Mean algorithm accuracy for the ring and tandem applicators was 0.60 ± 0.24 mm and 0.58 ± 0.24 mm, respectively.
• The algorithm's mean execution time was 10.0 seconds.

Conclusions:

• The developed algorithm's accuracy and speed are suitable for clinical use in applicator reconstruction.
• This automated approach represents a significant step towards fully automated brachytherapy treatment planning for cervical cancer.