A study on the identification of colonic polyp based on texture analysis of CT image

孟江,卢虹冰,徐肖攀,徐桓,张国鹏,梁正荣

DOI: https://doi.org/10.3969/J.ISSN.1672-8270.2018.03.003

2018-01-01

Abstract:Objective: To extract three dimensional (3D) Haralick characteristic from CT image of colon and its high-order partial derivative image and find out optimal characteristic so as to accurately differentiate polyp from normal wall tissue of colon. Methods: Totally 387 colonic polyps and 387 normal wall tissues of colon were extracted from the image data of CT colonoscope (CTC) of 111 patients with colonic polyp as 3D volumes of interest (VOIs) , and 3D gradient and curvature maps were calculated and transformed on these VOIs. And co-occurrence matrices were calculated from CT image, gradient image and curvature image of each VOI. And 60 Haralick textural features were extracted from each kind of co-occurrence matrices, and totally 180 features were extracted from each VOI. The feature selection strategy based on random forest was adopted to find out optimal feature. Finally, the optimal subset was input into four different classifiers, which included of logistic regression (LR), support vector machine (SVM), random forest (RF) and K-near neighbour (KNN), respectively, to validate its classification performance in differentiating polyps from wall tissues. Results: 37 features that selected from 180 features were confirmed as the optimal feature subset. And the optimal feature subset were validated in four different classifier and the results indicated that the average sensitivity of identifying polyp from wall tissue was above 99%, and the average specificity of that was above 97%, and all of the average area under the curve (AUC) of performance were 0.99. Conclusion: The textural feature that based on gray level, gray gradient and gray curvature has better specificity and sensitivity, and it can effectively identify polyp and normal wall tissue in the colon. Therefore, it has importantly referential value for clinical screening and diagnosis of colonic polyp based on imaging.