Prostate contours delineation using interactive directional active contours model and parametric shape prior model.

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Biomedical Engineering

Background:

• Accurate prostate contour delineation on Magnetic Resonance (MR) images is crucial for guiding biopsies, surgery, and therapy.
• Traditional active contour methods struggle with diffuse edges and complex anatomical structures near the prostate in MR images.
• Existing automated methods often lack the precision required for clinical applications due to these challenges.

Purpose of the Study:

• To develop a robust and accurate bi-stage algorithm for prostate contour delineation on MR images.
• To improve upon the limitations of traditional active contours in challenging imaging scenarios.
• To incorporate prior knowledge of prostate shape and user feedback for enhanced delineation accuracy.

Main Methods:

• Proposed a novel bi-stage delineation framework utilizing directional active contours (DAC).
• Integrated statistical and parametric shape prior models within a fast, globally acting DAC approach.
• Incorporated user feedback and a cost functional combining user input with shape priors for refinement.

Main Results:

• The proposed algorithm demonstrated effectiveness on publicly available prostate MR datasets, including challenging cases.
• The bi-stage approach successfully delineated prostate contours, overcoming limitations of standard methods.
• Comparison with state-of-the-art methods indicated the proposed algorithm's superior capability and effectiveness.

Conclusions:

• The novel bi-stage directional active contour algorithm provides an effective solution for prostate contour delineation in MR imaging.
• The integration of shape priors and user feedback significantly enhances delineation accuracy, especially in complex cases.
• This method holds promise for improving the precision of image-guided prostate interventions.