CAJ | 학술논문

基于增强计算机断层扫描图像，提出肾上腺肿瘤的计算机辅助诊断与分类算法。先读入肾上腺肿瘤的DICOM图像；然后利用基于局部区域的水平集方法分割肿瘤，解决边界模糊和区域不均匀的问题；在图像分割基础上，自动提取肿瘤的形态和纹理特征；最后利用支撑向量机进行分类并输出诊断结果(肾上腺皮质肿瘤或髓质肿瘤)。实验使用318幅增强计算机断层扫描图像，结果表明该算法具有良好的分类性能，其中准确性为95.28%、特异性为96.49%、敏感性为94.61%、阳性预测率为97.97%和阴性预测率为90.91%。因此，本文提出的计算机辅助诊断算法能够有效地对肾上腺肿瘤做出正确定位和定性诊断，可为临床的治疗和手术方案提供重要的参考。
기우증강계산궤단층소묘도상，제출신상선종류적계산궤보조진단여분류산법。선독입신상선종류적DICOM도상；연후이용기우국부구역적수평집방법분할종류，해결변계모호화구역불균균적문제；재도상분할기출상，자동제취종류적형태화문리특정；최후이용지탱향량궤진행분류병수출진단결과(신상선피질종류혹수질종류)。실험사용318폭증강계산궤단층소묘도상，결과표명해산법구유량호적분류성능，기중준학성위95.28%、특이성위96.49%、민감성위94.61%、양성예측솔위97.97%화음성예측솔위90.91%。인차，본문제출적계산궤보조진단산법능구유효지대신상선종류주출정학정위화정성진단，가위림상적치료화수술방안제공중요적삼고。
An algorithm based on contrast-enhanced computed tomography (CECT) images is proposed for the computer-aided diagnosis (CAD) and classification of adrenal masses. Firstly, images of adrenal masses are read in DICOM format. Then, a localized region-based level set method is utilized to handle the mass segmentation in cases with boundary ambiguity and inhomogeneity. Afterwards, morphological and texture features of masses are extracted based on segmented regions. Finally, the support vector ma-chine is used for classification to obtain the diagnosis result (adrenocortical or medullary mass). Experiments in 318 CECT images indicate the good classification performance of the proposed algorithm with the accuracy 95.28%, the specificity 96.49%, the sensi-tivity 94.61%, the positive predictive value 97.97% and the negative predictive value 90.91%. Hence, the proposed algorithm for CAD can efficiently and accurately locate and characterize adrenal masses which offers the significant reference for the clinical treatment and surgical planning.