Image segmentation errors correction by mesh segmentation and deformation.

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Image Segmentation

Background:

• Volumetric image segmentation frequently yields imperfect delineations of anatomical structures and pathologies.
• Manual correction of these segmentation errors is time-consuming and labor-intensive.

Purpose of the Study:

• To develop a semi-automatic method for efficient and accurate correction of volumetric image segmentation errors.
• To improve the precision of anatomical structure and pathology segmentation in medical imaging.

Main Methods:

• A novel semi-automatic approach utilizing min-cut segmentation to identify erroneous mesh vertices.
• Laplace deformation is applied to correct vertex coordinates based on local geometry.
• The method supports rapid user interaction on a 2D surface rendering.

Main Results:

• Demonstrated significant improvements in segmentation accuracy.
• Achieved an 83% reduction in average surface distance error.
• Reduced volume overlap error by 75% compared to initial segmentations.

Conclusions:

• The proposed method offers a robust and versatile solution for correcting segmentation errors across various anatomical structures and pathologies.
• Its independence from initial segmentation methods and fixed parameter values enhance its practical applicability.
• The technique significantly enhances the accuracy of volumetric image segmentation for clinical use.