研究目的
To devise a novel segmentation method for segmenting lesions in dermoscopy images using variational level sets formulation with a supervised learning aspect.
研究成果
The proposed method outperforms other state-of-the-art methods in segmenting dermoscopy images by incorporating a supervised learning aspect into the variational level sets formulation, leading to more accurate segmentation results.
研究不足
The method's sensitivity to the selection of an initial contour and the reliance on the edge indicator function which may exhibit weak lesion boundaries.
1：Experimental Design and Method Selection:
The methodology involves the use of variational level sets formulation with a novel area term based on supervised learning.
2：Sample Selection and Data Sources:
The PH2 dataset of 200 dermoscopy images was used, split into training and testing sets.
3：List of Experimental Equipment and Materials:
Not explicitly mentioned.
4：Experimental Procedures and Operational Workflow:
The method involves learning distributions for objects and backgrounds from training images, followed by curve fitting using an area term by minimizing the distance between distributions obtained from training and testing images.
5：Data Analysis Methods:
Performance was measured using Dice similarity coefficient (DSC) and F-Score.
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
