Texture analysis of CT images in predicting malignancy risk of gastrointestinal stromal tumours. Liu S,Pan X,Liu R,Zheng H,Chen L,Guan W,Wang H,Sun Y,Tang L,Guan Y,Ge Y,He J,Zhou Z Clinical radiology AIM:To explore the role of texture analysis of computed tomography (CT) images in predicting the malignancy risk of gastrointestinal stromal tumours (GISTs). MATERIALS AND METHODS:Seventy-eight patients with histopathologically confirmed GISTs underwent preoperative CT. Texture analysis was performed on unenhanced and contrast-enhanced CT images, respectively. Fourteen CT texture parameters were obtained and compared among GISTs at different malignancy risks with one-way analysis of variance or independent-samples Kruskal-Wallis test. Correlations between CT texture parameters and malignancy risk were analysed with Spearman's correlation test. Diagnostic performance of CT texture parameters in differentiating GISTs at low/very low malignancy risk was tested with receiver operating characteristic (ROC) analysis. RESULTS:Three parameters on unenhanced images (r=-0.268-0.506), four parameters on arterial phase (r=-0.365-0.508), and six parameters on venous phase (r=-0.343-0.481) imaging correlated significantly with malignancy risk of GISTs, respectively (all p<0.05). For identifying GISTs at low/very low malignancy risk, three parameters on unenhanced images (area under ROC curve [AUC], 0.676-0.802), four parameters on arterial phase (AUC, 0.637-0.811), and six parameters on venous phase (AUC, 0.636-0.791) imaging showed significant diagnostic performance, respectively (all p<0.05), especially maximum frequency on both unenhanced and contrast-enhanced images (AUC, 0.791-0.811). CONCLUSION:Texture analysis of CT images holds great potential to predict the malignancy risk of GISTs preoperatively. 10.1016/j.crad.2017.09.003

Machine learning-based quantitative texture analysis of CT images of small renal masses: Differentiation of angiomyolipoma without visible fat from renal cell carcinoma. Feng Zhichao,Rong Pengfei,Cao Peng,Zhou Qingyu,Zhu Wenwei,Yan Zhimin,Liu Qianyun,Wang Wei European radiology OBJECTIVE:To evaluate the diagnostic performance of machine-learning based quantitative texture analysis of CT images to differentiate small (≤ 4 cm) angiomyolipoma without visible fat (AMLwvf) from renal cell carcinoma (RCC). METHODS:This single-institutional retrospective study included 58 patients with pathologically proven small renal mass (17 in AMLwvf and 41 in RCC groups). Texture features were extracted from the largest possible tumorous regions of interest (ROIs) by manual segmentation in preoperative three-phase CT images. Interobserver reliability and the Mann-Whitney U test were applied to select features preliminarily. Then support vector machine with recursive feature elimination (SVM-RFE) and synthetic minority oversampling technique (SMOTE) were adopted to establish discriminative classifiers, and the performance of classifiers was assessed. RESULTS:Of the 42 extracted features, 16 candidate features showed significant intergroup differences (P < 0.05) and had good interobserver agreement. An optimal feature subset including 11 features was further selected by the SVM-RFE method. The SVM-RFE+SMOTE classifier achieved the best performance in discriminating between small AMLwvf and RCC, with the highest accuracy, sensitivity, specificity and AUC of 93.9 %, 87.8 %, 100 % and 0.955, respectively. CONCLUSION:Machine learning analysis of CT texture features can facilitate the accurate differentiation of small AMLwvf from RCC. KEY POINTS:• Although conventional CT is useful for diagnosis of SRMs, it has limitations. • Machine-learning based CT texture analysis facilitate differentiation of small AMLwvf from RCC. • The highest accuracy of SVM-RFE+SMOTE classifier reached 93.9 %. • Texture analysis combined with machine-learning methods might spare unnecessary surgery for AMLwvf. 10.1007/s00330-017-5118-z

Training-Based Gradient LBP Feature Models for Multiresolution Texture Classification. Luping Ji ,Yan Ren ,Guisong Liu ,Xiaorong Pu IEEE transactions on cybernetics Local binary pattern (LBP) is a simple, yet efficient coding model for extracting texture features. To improve texture classification, this paper designs a median sampling regulation, defines a group of gradient LBP (gLBP) descriptors, proposes a training-based feature model mapping method, and then develops a texture classification frame using the multiresolution feature fusion of four gLBP descriptors. Cooperated by median sampling, four descriptors encode a pixel respectively by central gradient, radial gradient, magnitude gradient and tangent gradient to generate initial gLBP patterns. The feature mapping models of gLBP descriptors are constructed by the maximal relative-variation rate (mr2) of rotation-invariant patterns, and then prestored as mapping lookup files. By mapping, initial patterns can be transformed into low-dimensional ones. And then it generates multiresolution texture features via the joint and concatenation of gLBP descriptors on different sampling parameters. A trained nearest neighbor classifier with chi-square distance is applied to classify textures by feature histograms. The experimental results of simulation on five public texture databases show that the proposed method is reliable and efficient in texture classification. In comparison with nine other similar approaches, including two state-of-the-art ones, the proposed method runs faster than most of them and also outperforms all of them in terms of classification accuracy and noise robustness. It achieves higher accuracy and has also better robustness to the Salt&Pepper and Gaussian noise added artificially into texture images. 10.1109/TCYB.2017.2748500