Related Experiment Videos
Computer-aided kidney segmentation on abdominal CT images
Daw-Tung Lin1, Chung-Chih Lei, Siu-Wan Hung
1Department of Computer Science and Information Engineering, National Taipei University, Taiwan, ROC. dalton@mail.ntpu.edu.tw
Summary
This study introduces an automatic kidney segmentation model for abdominal CT scans, achieving 88% correlation with manual segmentation. This computer-aided tool aids physicians in clinical diagnosis and medical education.
Area of Science:
- Medical Imaging
- Computer-Aided Diagnosis
- Anatomical Segmentation
Background:
- Accurate kidney segmentation is crucial for clinical diagnosis and medical training.
- Existing segmentation methods may lack robustness in handling anatomical variations and pathologies.
Purpose of the Study:
- To develop an effective model-based approach for automatic kidney segmentation in abdominal CT images.
- To create a system that considers anatomic structures for improved accuracy and clinical utility.
Main Methods:
- A two-stage coarse-to-fine segmentation approach utilizing the spine as a landmark.
- Elliptic candidate region extraction with progressive positioning and a novel directional model for seed point identification.
- Adaptive region growing controlled by image homogeneity and a visualization tool for renal contour display.
Main Results:
- The system demonstrated an average correlation coefficient of 88% between automatic and manual segmentation.
- The method was tested on 358 images from 30 patients, including those with pathologies.
- The approach is adaptable to images of varying sizes using relative kidney-to-spine distances.
Conclusions:
- The proposed model-based approach offers an effective and automatic solution for kidney segmentation in abdominal CT scans.
- The system shows promise in assisting physicians with clinical diagnosis and enhancing educational training.
- Consideration of anatomic structures and pathologies contributes to the reliability of the segmentation results.