Texture feature based liver lesion classification

Yeela Doron; Nitzan Mayer-Wolf; Idit Diamant; Hayit Greenspan

doi:10.1117/12.2043697

24 March 2014 Texture feature based liver lesion classification

Yeela Doron, Nitzan Mayer-Wolf, Idit Diamant, Hayit Greenspan

Proceedings Volume 9035, Medical Imaging 2014: Computer-Aided Diagnosis; 90353K (2014) https://doi.org/10.1117/12.2043697
Event: SPIE Medical Imaging, 2014, San Diego, California, United States

Abstract

Liver lesion classification is a difficult clinical task. Computerized analysis can support clinical workflow by enabling more objective and reproducible evaluation. In this paper, we evaluate the contribution of several types of texture features for a computer-aided diagnostic (CAD) system which automatically classifies liver lesions from CT images. Based on the assumption that liver lesions of various classes differ in their texture characteristics, a variety of texture features were examined as lesion descriptors. Although texture features are often used for this task, there is currently a lack of detailed research focusing on the comparison across different texture features, or their combinations, on a given dataset. In this work we investigated the performance of Gray Level Co-occurrence Matrix (GLCM), Local Binary Patterns (LBP), Gabor, gray level intensity values and Gabor-based LBP (GLBP), where the features are obtained from a given lesion`s region of interest (ROI). For the classification module, SVM and KNN classifiers were examined. Using a single type of texture feature, best result of 91% accuracy, was obtained with Gabor filtering and SVM classification. Combination of Gabor, LBP and Intensity features improved the results to a final accuracy of 97%.

Citation Download Citation

Yeela Doron, Nitzan Mayer-Wolf, Idit Diamant, and Hayit Greenspan "Texture feature based liver lesion classification", Proc. SPIE 9035, Medical Imaging 2014: Computer-Aided Diagnosis, 90353K (24 March 2014); https://doi.org/10.1117/12.2043697

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available