研究目的
To propose a new adaptive and non-parametric segmentation method called Optimum Path Snakes (OPS) for segmenting medical CT images of the lung and brain, with automatic initialization and stop criteria, aiming to improve accuracy and efficiency compared to existing methods.
研究成果
The OPS method is a promising tool for medical image segmentation, providing high accuracy (DC close to 1) and low processing times in many cases. It outperforms or matches specialized methods like SISDEP and CRAD for lung segmentation and shows robustness to noise. Future work includes expansion to color images and other applications.
研究不足
The method requires preprocessing for feature extraction, and performance may vary with different image types and noise levels. It is computationally intensive for some configurations, e.g., with GLCM extractor and larger masks.
1：Experimental Design and Method Selection:
The OPS method is based on active contour models, incorporating internal energy (Optimum Balloon Adaptive) and external energy (Optimum Path) derived from graph theory and the Optimum Path Forest classifier. It uses automatic initialization and stop criteria.
2：Sample Selection and Data Sources:
Lung CT images from 36 patients (12 healthy, 12 with fibrosis, 12 with COPD) and brain CT images from 100 patients with stroke, acquired using Toshiba Aquilion, GE Medical System LightSpeed16, and Philips Brilliance10 tomographs.
3：List of Experimental Equipment and Materials:
CT scanners (Toshiba Aquilion, GE Medical System LightSpeed16, Philips Brilliance10), DICOM images, DCMTK library for image reading, computer with Windows 7, Intel Core i5 processor, 3.20 GHz, 8GB RAM.
4：20 GHz, 8GB RAM.
Experimental Procedures and Operational Workflow:
4. Experimental Procedures and Operational Workflow: Images are preprocessed using texture analysis (GLCM, HuMom, StatMom, AHTD). OPS is applied with various configurations (Euclidean, Gaussian, Manhattan distances; mask sizes 3x3 to 15x15). Segmentation results are compared to manual segmentations by specialists using Dice coefficient and Hausdorff distance metrics.
5：5). Segmentation results are compared to manual segmentations by specialists using Dice coefficient and Hausdorff distance metrics.
Data Analysis Methods:
5. Data Analysis Methods: Statistical analysis using Dice coefficient and Hausdorff distance to evaluate segmentation accuracy and processing time.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
