研究目的
To automate the identification of antigen patterns in HEp-2 cell images for the diagnosis of auto-immune diseases.
研究成果
The proposed framework, which combines Laws features with two level SVM classifier, yields high classification accuracies on two publicly available databases, demonstrating its effectiveness in automating the identification of antigen patterns in HEp-2 cell images.
研究不足
The study is limited by the availability of a constrained form of stained HEp-2 cell images, which may not cover all possible variations in cell patterns.
1：Experimental Design and Method Selection:
The framework combines Laws features with two level SVM classifier coupled with posterior class probabilities during the classification stage.
2：Sample Selection and Data Sources:
The MIVIA ICPR 2012 Contest Dataset and the SNP HEp-2 Cell Dataset are used.
3：List of Experimental Equipment and Materials:
Fluorescence microscope coupled with a mercury vapor lamp and a digital camera for MIVIA dataset; monochrome high dynamic range microscopy camera and an LED illumination source for SNP dataset.
4：Experimental Procedures and Operational Workflow:
Individual cell images are extracted and used to extract efficient representation followed by classification.
5：Data Analysis Methods:
The performance is evaluated using leave-one-out technique over all images in the dataset for MIVIA and pre-defined five-fold validation training and testing splits for SNP.
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
