Related Experiment Videos
Colonic polyp segmentation in CT colonography-based on fuzzy clustering and deformable models
Jianhua Yao1, Meghan Miller, Marek Franaszek
1Diagnostic Radiology Department, Clinical Center, the National Institutes of Health, 10 Center Drive, MSC 1182, Bethesda, MD 20892-1182 USA.
IEEE Transactions on Medical Imaging
|November 24, 2004
Summary
This study introduces an automated method for segmenting colonic polyps in CT colonography. The computer-aided detection system achieves repeatability comparable to human experts and significantly reduces false positives.
Area of Science:
- Medical Imaging
- Computer-Aided Diagnosis
- Gastroenterology
Background:
- Accurate segmentation of colonic polyps in computed tomography (CT) colonography is crucial for early cancer detection.
- Manual segmentation is time-consuming and subject to inter-observer variability.
- Developing automated methods can improve efficiency and consistency in polyp detection.
Purpose of the Study:
- To present an automatic method for segmenting colonic polyps using CT colonography.
- To validate the accuracy and repeatability of the automated segmentation method.
- To assess the utility of the segmentation in a computer-aided detection (CAD) system for reducing false positives and aiding polyp classification.
Main Methods:
- A novel automatic segmentation method combining knowledge-guided intensity adjustment, fuzzy c-mean clustering, and deformable models.
- Validation through comparison of computer-generated segmentations with manual segmentations.
- Investigation of intraoperator and interoperator repeatability for manual segmentation.
- Application of segmentation in a CAD system to reduce false positives and extract volumetric features for classification.
Main Results:
- An average 76.3% volume overlap percentage was achieved in validating 105 polyp detections, demonstrating good accuracy for small polyps.
- Computer-human segmentation repeatability was found to be comparable to interoperator repeatability.
- The segmentation method eliminated 30% of false positive detections when applied in CAD.
- Volumetric features derived from segmentation further reduced false positives by 50% at 80% sensitivity.
Conclusions:
- The proposed automatic method provides accurate and repeatable colonic polyp segmentation in CT colonography.
- The system demonstrates significant potential for improving the efficiency and accuracy of polyp detection and classification in clinical practice.
- Automated segmentation is a valuable tool for reducing radiologist workload and enhancing diagnostic performance in colorectal cancer screening.