Application of Radiomics Based on Enhanced Chest CT Scan in Differential
Diagnosis of Thymoma
MO Zhanhao1，SUI He1，LI Xuejia2，SUN Xu3，LIU Jingxin1
1. Department of Radiology, China-Japan Union Hospital, Jilin University, Changchun Jilin 130033, China|
2. Department of Radiology, Qingdao Municipal Hospital, Qingdao Shandong 266011, China|
3. Department of Radiology, The Second Hospital of Jilin University, Changchun Jilin 130000, China
Abstract：Objective To retrospectively analyze the radiomic characteristics of venous phase contrast enhanced CT (CECT) scan to distinguish thymoma from other anterior mediastinal lesions, so as to provide more for preoperative diagnosis, treatment and prognostic evaluation of thymoma reference information. Methods A total of 168 patients with anterior mediastinal disease with detailed pathological information were enrolled. All patients received chest CECT before treatment or before surgery. A total of 1029 radiomic features were extracted from the venous CECT image, and the variance threshold, the optimal K value, and the least absolute contraction and selection operator algorithm were used in turn to screen the radiomic feature values. Radiomic characteristics should be the most relevant to distinguish thymoma and other anterior mediastinal space-occupying lesions. Logistic regression was used as a feature selection classi?er for normalized data, thereby constructing a radiomic model composed of 19 relevant omics features. Finally, the receiver operating characteristic curve was used to evaluate the predictive classi?cation performance of the radiomics model. Results Among the 1029 radioomic features extracted, 19 features had signi?cant differences in distinguishing thymoma from other anterior mediastinal lesions, including Square root_glszm_Zone Entropy, Wavelet-LLH_glrlm_Run Entropy, wavelet-LHL_ ?rst order_Maximum, etc. The accuracy of using this radiomic model to distinguish thymoma from other anterior mediastinal lesions was 85%, the sensitivity was 88%, and the speci?city was 82%. Conclusion The radiomic model based on venous thoracic CECT has good correlation and accuracy in the identi?cation of thymoma and other anterior mediastinal lesions, and has a high application potential in the clinical diagnosis of thymoma.
莫展豪1，隋赫1，李雪佳2，孙旭3，刘景鑫1. 基于胸部增强CT扫描的放射组学方法在胸腺瘤鉴别诊断中的应用[J]. 中国医疗设备, 2020, 35(10): 23-27.
MO Zhanhao1，SUI He1，LI Xuejia2，SUN Xu3，LIU Jingxin1. Application of Radiomics Based on Enhanced Chest CT Scan in Differential
Diagnosis of Thymoma. China Medical Devices, 2020, 35(10): 23-27.
Carter BW,Marom EM,Detterbeck FC.Approaching the patient with an anterior mediastinal mass: a guide for clinicians[J]. J Thorac Oncol,2014,9(9 S2):S102-S109.
