Improving Cell Image Classification with Curvelet and Haralick Features
Author Information
Author(s): Zhang Bailing, Pham Tuan D
Primary Institution: Xi'an Jiaotong-Liverpool University
Hypothesis
Can combining curvelet transform and Haralick features improve the classification accuracy of cellular phenotypes in microscopy images?
Conclusion
The study demonstrates that using a random subspace ensemble of classifiers with combined curvelet and Haralick features significantly improves classification accuracy for microscopy images.
Supporting Evidence
- The random subspace ensemble achieved classification rates of 91.20%, 98.86%, and 91.03% for HeLa, CHO, and RNAi image sets respectively.
- The combined feature approach outperformed individual feature methods in classification accuracy.
- The study found that curvelet features are superior to wavelet features for bioimage descriptions.
Takeaway
This study shows that by using special techniques to analyze images of cells, we can better tell what type of cells they are, which helps scientists understand them more easily.
Methodology
The study used a random subspace ensemble classifier with multiple layer perceptrons, combining features from curvelet transform and Haralick features to classify microscopy images.
Potential Biases
Potential biases may arise from the selection of image datasets and the specific feature extraction methods employed.
Limitations
The study may be limited by the specific datasets used and the generalizability of the results to other types of images.
Statistical Information
P-Value
p<0.05
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website