Fully Automatic Segmentation and Three-Dimensional Reconstruction of the Liver in CT Images

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Biomedical Engineering

Background:

• Accurate liver segmentation is crucial for diagnosing liver diseases and planning surgical interventions.
• The variability in liver shape and intensity, along with its connection to other organs across CT slices, presents significant challenges for automated segmentation.

Purpose of the Study:

• To develop an effective and robust automatic method for liver segmentation and 3D reconstruction from CT images.
• To overcome the difficulties posed by complex anatomical variations and intensity similarities between the liver and adjacent organs.

Main Methods:

• A heuristic approach utilizing multiple thresholds computed from slope difference distribution for initial organ segmentation.
• Combining segmentation results from different thresholds to robustly delineate the liver boundary across all CT slices.
• Employing a global energy minimization function to reconstruct the 3D liver shape from identified 2D boundaries.

Main Results:

• The proposed method successfully segmented the liver automatically and robustly across various CT images.
• Experimental results demonstrated the effectiveness of the image processing algorithms in achieving accurate liver segmentation.
• The 3D reconstruction using global energy minimization proved effective in creating a coherent liver model.

Conclusions:

• The developed heuristic approach offers a reliable solution for automatic liver segmentation and 3D reconstruction in CT imaging.
• This technique has the potential to significantly improve the accuracy and efficiency of liver disease diagnosis and surgical planning.