Limits...
3D texture analysis in renal cell carcinoma tissue image grading.

Kim TY, Cho NH, Jeong GB, Bengtsson E, Choi HK - Comput Math Methods Med (2014)

Bottom Line: Next, we extracted quantitative values using a 3D gray level cooccurrence matrix (GLCM) and a 3D wavelet based on two types of basis functions.In the grade classification results, 3D Haar wavelet texture features combined with principal component analysis showed the best discrimination results.Classification using 3D wavelet texture features was significantly better than 3D GLCM, suggesting that the former has potential for use in a computer-based grading system.

View Article: PubMed Central - PubMed

Affiliation: Department of Computer Engineering, Inje University, Injero 197, UHRC, Gimhae, Gyeongnam 621-749, Republic of Korea.

ABSTRACT
One of the most significant processes in cancer cell and tissue image analysis is the efficient extraction of features for grading purposes. This research applied two types of three-dimensional texture analysis methods to the extraction of feature values from renal cell carcinoma tissue images, and then evaluated the validity of the methods statistically through grade classification. First, we used a confocal laser scanning microscope to obtain image slices of four grades of renal cell carcinoma, which were then reconstructed into 3D volumes. Next, we extracted quantitative values using a 3D gray level cooccurrence matrix (GLCM) and a 3D wavelet based on two types of basis functions. To evaluate their validity, we predefined 6 different statistical classifiers and applied these to the extracted feature sets. In the grade classification results, 3D Haar wavelet texture features combined with principal component analysis showed the best discrimination results. Classification using 3D wavelet texture features was significantly better than 3D GLCM, suggesting that the former has potential for use in a computer-based grading system.

Show MeSH

Related in: MedlinePlus

Comparison of classification performance using 6 different models.
© Copyright Policy - open-access
Related In: Results  -  Collection


getmorefigures.php?uid=PMC4209774&req=5

fig6: Comparison of classification performance using 6 different models.

Mentions: Classifier F, which applied PCA, showed similar results. Its classification accuracy was 25.00% for grade 1, 54.55% for grade 2, 50.00% for grade 3, and 63.64% for grade 4. The overall classification accuracy was 53.13%. Again, the classification of grade 1 data was generally incorrect. Both classifiers showed high classification accuracy for the training data, but the results were lower than 50% for the test data. This may be caused by the fact that the data extraction allows many degrees of freedom, which results in overtraining on the training data because of the low number of data sets we have available. Figure 6 compares the classification results for the six models using the test data. The statistical analysis results indicated that a considerable amount of data was misclassified into grade 2. However, classifier B solved a considerable portion of this problem and showed the most stable accuracy when compared to the other five classifiers. The results were confirmed by 2 pathologists and the correlation study between the subjective and computerized grading.


3D texture analysis in renal cell carcinoma tissue image grading.

Kim TY, Cho NH, Jeong GB, Bengtsson E, Choi HK - Comput Math Methods Med (2014)

Comparison of classification performance using 6 different models.
© Copyright Policy - open-access
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC4209774&req=5

fig6: Comparison of classification performance using 6 different models.
Mentions: Classifier F, which applied PCA, showed similar results. Its classification accuracy was 25.00% for grade 1, 54.55% for grade 2, 50.00% for grade 3, and 63.64% for grade 4. The overall classification accuracy was 53.13%. Again, the classification of grade 1 data was generally incorrect. Both classifiers showed high classification accuracy for the training data, but the results were lower than 50% for the test data. This may be caused by the fact that the data extraction allows many degrees of freedom, which results in overtraining on the training data because of the low number of data sets we have available. Figure 6 compares the classification results for the six models using the test data. The statistical analysis results indicated that a considerable amount of data was misclassified into grade 2. However, classifier B solved a considerable portion of this problem and showed the most stable accuracy when compared to the other five classifiers. The results were confirmed by 2 pathologists and the correlation study between the subjective and computerized grading.

Bottom Line: Next, we extracted quantitative values using a 3D gray level cooccurrence matrix (GLCM) and a 3D wavelet based on two types of basis functions.In the grade classification results, 3D Haar wavelet texture features combined with principal component analysis showed the best discrimination results.Classification using 3D wavelet texture features was significantly better than 3D GLCM, suggesting that the former has potential for use in a computer-based grading system.

View Article: PubMed Central - PubMed

Affiliation: Department of Computer Engineering, Inje University, Injero 197, UHRC, Gimhae, Gyeongnam 621-749, Republic of Korea.

ABSTRACT
One of the most significant processes in cancer cell and tissue image analysis is the efficient extraction of features for grading purposes. This research applied two types of three-dimensional texture analysis methods to the extraction of feature values from renal cell carcinoma tissue images, and then evaluated the validity of the methods statistically through grade classification. First, we used a confocal laser scanning microscope to obtain image slices of four grades of renal cell carcinoma, which were then reconstructed into 3D volumes. Next, we extracted quantitative values using a 3D gray level cooccurrence matrix (GLCM) and a 3D wavelet based on two types of basis functions. To evaluate their validity, we predefined 6 different statistical classifiers and applied these to the extracted feature sets. In the grade classification results, 3D Haar wavelet texture features combined with principal component analysis showed the best discrimination results. Classification using 3D wavelet texture features was significantly better than 3D GLCM, suggesting that the former has potential for use in a computer-based grading system.

Show MeSH
Related in: MedlinePlus