Texture Analysis of Supraspinatus Ultrasound Image for Computer Aided Diagnostic System
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
Objectives: In this paper, we proposed an algorithm for recognizing a rotator cuff supraspinatus tendon tear using a texture analysis based on a histogram, gray level co-occurrence matrix (GLCM), and gray level run length matrix (GLRLM). Methods: First, we applied a total of 57 features (5 first order descriptors, 40 GLCM features, and 12 GLRLM features) to each rotator cuff region of interest. Our results show that first order statistics (mean, skewness, entropy, energy, smoothness), GLCM (correlation, contrast, energy, entropy, difference entropy, homogeneity, maximum probability, sum average, sum entropy), and GLRLM features are helpful to distinguish a normal supraspinatus tendon and an abnormal supraspinatus tendon. The statistical significance of these features is verified using a t-test. The support vector machine classification showed accuracy using feature combinations. Support Vector Machine offers good performance with a small amount of training data. Sensitivity, specificity, and accuracy are used to evaluate performance of a classification test. Results: From the results, first order statics features and GLCM and GLRLM features afford 95%, 85%, and 100% accuracy, respectively. First order statistics and GLCM and GLRLM features in combination provided 100% accuracy. Combinations that include GLRLM features had high accuracy. GLRLM features were confirmed as highly accurate features for classified normal and abnormal. Conclusions: This algorithm will be helpful to diagnose supraspinatus tendon tear on ultrasound images.
Abstract
Objectives: In this paper, we proposed an algorithm for recognizing a rotator cuff supraspinatus tendon tear using a texture analysis based on a histogram, gray level co-occurrence matrix (GLCM), and gray level run length matrix (GLRLM). Methods: First, we applied a total of 57 features (5 first order descriptors, 40 GLCM features, and 12 GLRLM features) to each rotator cuff region of interest. Our results show that first order statistics (mean, skewness, entropy, energy, smoothness), GLCM (correlation, contrast, energy, entropy, difference entropy, homogeneity, maximum probability, sum average, sum entropy), and GLRLM features are helpful to distinguish a normal supraspinatus tendon and an abnormal supraspinatus tendon. The statistical significance of these features is verified using a t-test. The support vector machine classification showed accuracy using feature combinations. Support Vector Machine offers good performance with a small amount of training data. Sensitivity, specificity, and accuracy are used to evaluate performance of a classification test. Results: From the results, first order statics features and GLCM and GLRLM features afford 95%, 85%, and 100% accuracy, respectively. First order statistics and GLCM and GLRLM features in combination provided 100% accuracy. Combinations that include GLRLM features had high accuracy. GLRLM features were confirmed as highly accurate features for classified normal and abnormal. Conclusions: This algorithm will be helpful to diagnose supraspinatus tendon tear on ultrasound images.
참고문헌 (11)
- Chakravarty K, "Shoulder joint movement and its relationship to disability in the elderly", J Rheumatol, 20(8), pp.1359~1361 (1993)
- Chen WM, "3-D ultrasound texture classification using run difference matrix", Ultrasound Med Biol, 31(6), pp.763~770 (2005)
- Horng MH, "Texture classification of the ultrasonic images of rotator cuff diseases based on radial basis function network", Proceedings of 2008 IEEE International Joint Conference on Neural Networks, pp.91~97 (2008)
- "Supraspinatus muscle" (2016)[URL]
- Comparative effectiveness research of conservative treatment and rotator cuff repair for the patient with rotator cuff tears, "Lim JY", National Evidence-based Healthcare Collaborating Agency, ~ (2015)
- Haralick RM, "Textural features for image classification", IEEE Trans Syst Man Cybern, 3(6), pp.610~621 (1973)
- Galloway MM, "Texture analysis using gray level run lengths", Comput Graph Image Process, 4(2), pp.172~179 (1975)
- Mavroforakis ME, "Mammographic masses characterization based on localized texture and dataset fractal analysis using linear, neural and support vector machine classifiers", Artif Intell Med, 37(2), pp.145~162 (2006)
- Chen YW, "Feature extraction", pp.315~324 (2006), Springer
- "Sensitivity and specificity" (2016)[URL]
- Andarawis-Puri N, "Rotator cuff tendon strain correlates with tear propagation", J Biomech, 42(2), pp.158~163 (2009)