|
|
The Classification of Pulmonary Nodules Based on Texture Features over Local Jet Transformation Space |
Dai Meiling1, Qi Jin1,3, Zhou Zhongxing1,2, Gao Feng 1,2#* |
1 (College of Precision Instrument and Optoelectronics Engineering,Tianjin University,Tianjin 300072,China) 2(Tianjin Key Laboratory of Biomedical Detecting Techniques and Instruments,Tianjin 300072,China) 3(Cancer Institute and Hospital, Tianjin Medical University, Tianjin 300060, China) |
|
|
Abstract A method of texture features extraction over local jet transformation space is presented in order to improve the recognition about benign/malignant nodules with texture features. In this method, the first three order partial differential functions of 2-D Gaussian were used to transform original image of nodule tolocal jet texture images space firstly, and then feature parameters were extracted over the local jet images space by texture descriptor. In our work, feature parameters were extracted from the original CT image and local jet images using the texture descriptors of first four moments of intensity and parameters based on gray level co-occurrence matrix. Then BP neural network was used to classify the benign/malignant nodules with the two texture features sets under the same texture descriptor that optimized by kernel principle component analysis. The compared experiment conducted based on 157 lung nodule(51 benign and 106 malignant) samples. The experiment result showed that the classification accuracy of two feature parameters sets extracted by the two texture descriptors on local jet space was 82.69% and 86.54% respectively, higher 6%~8% than feature parameters sets on original image space, and the value of AUC was increased about 10%. The experiment result indicated the texture features extracted from local jet transformation space could effectively improve the pattern recognition of benign/malignant of pulmonary nodules.
|
Received: 22 April 2016
|
|
About author:: (Senior member, Chinese Society of Biomedical Engineering) |
|
|
|
[1] Koenderink J,VanDoom A. Representation of local geometry in the visual system[J]. Biological Cybernetics,1987,55(6):367-375. [2] William M,Silva D,Manzanera A,et al. Feature extraction on Local Jet space for texture classification[J]. Physica A: Statistical Mechanics & Its Application,2015,439(1):160-170. [3] Schmid C,Mohr R. Local gray-value invariants for image retrieval[J]. IEEE Transactions on Pattern Analysis & Machine, 1997,19(5):447-460. [4] Crosier M,Griffin L. Using basic image features for texture classification[J]. International Journal of Computer Vision, 2010, 88(3):447-460. [5] Manzanera A. Local Jet feature space framework for image processing and representation[C]//2011 Seventh International Conference on Signal Image Technology & Internet-Based System. Pennsylvania:HAL CCSD,2011:261-268. [6] Hogeweg L, Clara IS, Maduskar P, et al. Automatic detection of tuberculosis in chest radiographs using a combination of textural, focal, and shape abnormality analysis[J]. IEEE Transaction on Medical Imaging,2015,34(12):2429-2442. [7] Martinez F, Acosta O, Crevoisier RD, et al. Local Jet features and statistical models in a hybrid Bayesian framework for prostate estimation in CBCT image[J]. Proc SPIE,2012,8315(7):527-540. [8] Kuruvilla J, Gunavathi K. Lung cancer classification using neural networks for CT images[J]. Computer Methods & Programs in Biomedicine,2014,113(1):201-209. [9] Seongjin P, Bohyoung K, Jeongjin Lee, et al. GGO nodule volume-preserving non-rigid lung registration using GLCM texture analysis[J]. IEEE Transactions on Biomedical Engineering, 2011,58(10):2885-2894. [10] Balagurunathan Y, Yoganand, Gu Yuhua, et al. Reproducibility and prognosis of quantitative features extracted from CT image[J]. Translational Oncology,2014,7(1):72-87. [11] Kaya A, Can AB. A weighted rule based method for predicting malignancy of pulmonary nodules by nodule characteristics[J]. Journal of Biomedical Informatics,2015,56(8):69-79. |
|
|
|