Classifying Cyst and Tumor Lesion Using Support Vector Machine Based on Dental Panoramic Images Texture Features
Classifying Cyst and Tumor Lesion Using Support Vector Machine Based on Dental Panoramic Images Texture Features
复制标题
基于牙科全景图像纹理特征的支持向量机对囊肿和肿瘤病变进行分类
DOI：
--
复制
发表时间：
2013
期刊：
影响因子：
--
通讯作者：
M. Hariadi
中科院分区：
文献类型：
--
作者：
Ingrid Nurtanio;E. Astuti;K. Purnama;M. Hariadi
牙科X光片在诊断颌骨的病理学方面是必不可少的。然而，颌骨病变相似的影像学表现导致难以区分囊肿和肿瘤。因此，我们进行了计算机辅助分类系统的开发，在牙科全景图像囊肿和肿瘤病变。该系统包括基于纹理的特征提取，使用一阶统计纹理（FO），灰度共生矩阵（GLCM）和灰度游程长度矩阵（GLRLM）。在这项工作中，有33个功能，使用支持向量机（SVM）的分类。结果表明，囊肿与肿瘤病变的鉴别准确率可达87.18%，受试者工作特征曲线（AUC）下的面积可达0.9444。当使用特征数量作为预测因子时，获得的最高准确度为：使用FO为84.62%，使用GLCM为61.54%，使用GLRLM为76.92%，使用FO和GLCM的组合为84.62%，使用FO和GLRLM的组合为87.18%，使用GLCM和GLRLM的组合为75.56%，使用FO的组合为87.18%，GLCM和GLRLM。最高AUC值为FO为0.9361，GLCM为0.8667，GLRLM为0.8722，FO和GLCM联合为0.9278，FO和GLRLM联合为0.9444，GLCM和GLRLM联合为0.8417，FO、GLCM和GLRLM联合为0.9278。基于AUC值，该预测的准确度水平可以被归类为“不准确”。
Abstract— Dental radiographs are essential in diagnosing the pathology of the jaw. However, similar radiographic appearance of jaw lesions causes difficulties in differentiating cyst from tumor. Therefore, we conducted a development of computer-aided classification system for cyst and tumor lesions in dental panoramic images. The proposed system consists of feature extraction based on texture using the first-order statistics texture (FO), Gray Level Co-occurrence Matrix (GLCM) and Gray Level Run Length Matrix (GLRLM). In this work, there were thirty three features which were classified using Support Vector Machine (SVM) based classification. The result shows that differentiation of cyst from tumor lesions can achieve accuracy up to 87.18% and Area Under the Receiver Operating Characteristic (AUC) curve up to 0.9444. When using the number of features used as predictors, the highest accuracy obtained were 8462% using FO, 61.54% using GLCM, 76.92% using GLRLM, 84.62% using the combination of FO and GLCM, 87.18% using the combination of FO and GLRLM, 75.56% using the combination of GLCM and GLRLM, and 87.18% using the combination of FO, GLCM and GLRLM. The highest AUC value was 0.9361 using FO, using GLCM was 0.8667, using GLRLM was 0.8722, using the combination of FO and GLCM was 0.9278, using the combination of FO and GLRLM was 0.9444, using the combination of GLCM and GLRLM was 0.8417, and using the combination of FO, GLCM and GLRLM was 0.9278. Based on the AUC value, the level of accuracy of this prediction can be categorized as 'Excellent'.