A fully automatic 2D segmentation method for uterine fibroid in MRgFUS treatment evaluation

Area of Science:

• Medical Imaging
• Computational Anatomy
• Non-invasive Surgery

Background:

• Magnetic Resonance guided Focused UltraSound (MRgFUS) is a non-invasive treatment for uterine fibroids using thermal ablation.
• Manual segmentation of treated fibroid areas for NonPerfused Volume (NPV) evaluation is operator-dependent, time-consuming, and impacts reproducibility.
• Current manual segmentation methods can lead to errors in follow-up assessments and inefficient use of resources.

Purpose of the Study:

• To develop a fully automatic method for uterine fibroid segmentation post-MRgFUS treatment.
• To enhance the accuracy and reproducibility of NonPerfused Volume (NPV) evaluation.
• To reduce the time and operator dependency associated with manual segmentation.

Main Methods:

• A novel method employing unsupervised Fuzzy C-Means clustering and iterative optimal threshold selection.
• Segmentation of uterus and fibroid regions for post-operative evaluation.
• Testing on 15 MR datasets from patients with uterine fibroids.

Main Results:

• The automatic method achieved high segmentation accuracy with metrics such as SDI=88.67% and JI=80.70%.
• Quantitative evaluation using area-based and distance-based metrics demonstrated effectiveness.
• Results were compared favorably against similar literature approaches.

Conclusions:

• The proposed method offers a practical, automatic solution for evaluating ablated fibroid boundaries and volumes.
• No external user input is required, ensuring a fully automated process.
• The segmentation results validate the method's effectiveness and practicality for clinical use.