Statistical comparison of color model-classifier pairs in hematoxylin and eosin stained histological images

Authors:
Mutlu Mete;Umit Topaloglu
Affiliations:
Information Technology, University of Arkansas for Medical Sciences, Little Rock, AR;Information Technology, University of Arkansas for Medical Sciences, Little Rock, AR
Venue:
CIBCB'09 Proceedings of the 6th Annual IEEE conference on Computational Intelligence in Bioinformatics and Computational Biology
Year:
2009

Citing 12
Cited 0

C4.5: programs for machine learning

C4.5: programs for machine learning
The nature of statistical learning theory

The nature of statistical learning theory
Information Retrieval

Information Retrieval
Inference for the Generalization Error

Machine Learning
Comparison of Five Color Models in Skin Pixel Classification

RATFG-RTS '99 Proceedings of the International Workshop on Recognition, Analysis, and Tracking of Faces and Gestures in Real-Time Systems
Automatic Color Space Selection for Biological Image Segmentation

ICPR '04 Proceedings of the Pattern Recognition, 17th International Conference on (ICPR'04) Volume 3 - Volume 03
Digital Image Processing (3rd Edition)

Digital Image Processing (3rd Edition)
Data Mining: Practical Machine Learning Tools and Techniques, Second Edition (Morgan Kaufmann Series in Data Management Systems)

Data Mining: Practical Machine Learning Tools and Techniques, Second Edition (Morgan Kaufmann Series in Data Management Systems)
Head and Neck Cancer Detection in Histopathological Slides

ICDMW '06 Proceedings of the Sixth IEEE International Conference on Data Mining - Workshops
Histopathological Image Analysis Using Model-Based Intermediate Representations and Color Texture: Follicular Lymphoma Grading

Journal of Signal Processing Systems
Delineation of malignant areas in histological images of head-neck cancer

Delineation of malignant areas in histological images of head-neck cancer
Effect of colorspace transformation, the illuminance component, and color modeling on skin detection

CVPR'04 Proceedings of the 2004 IEEE computer society conference on Computer vision and pattern recognition

Quantified Score

Hi-index	0.01

Visualization

Abstract

Color is the most critical information for assessing histological images. However, in literature, there is no standard color space in which a particular color points are represented for computer vision tasks. In this paper, we evaluated 11 color models with three different learning schemas for their performance in classifying tumor-related colors. The color models we studied are CIELAB, CIELUV, CIEXYZ, CMY, CMYK, HSL, HSV, Hunter-LAB, NRGB, RGB, and SCT. With 11 color models, prediction accuracies of three well-known classifiers, namely SVMs, C4.5, and Naïve Bayes, are statistically compared on a large dataset of 3494 Hematoxylin and Eosin (HE) stained histopathologic images. Surprisingly, experiment results show that in contrast to general assumptions, there is no single model that is better than others in every case. However, C4.5 outperformed other two classifiers by obtaining average F-measure of 0.9989. Of 11 color models, we suggest the pair of C4.5-SCT as the most accurate classification framework for tumor identification in HE stained histological images.