Wen-Lei Qian1, Yu Jiang1, Xi Liu2, Ying-Kun Guo3, Yuan Li1, Xin Tang1, Zhi-Gang Yang4. 1. Department of Radiology, West China Hospital, Sichuan University, 37# Guo Xue Xiang, Chengdu, 610041, Sichuan, China. 2. Key Laboratory of Carcinogenesis and Translational Research (Ministry of Education), Department of Radiology, Peking University Cancer Hospital and Institute, No. 52 Fu Cheng Road Hai Dian District, Beijing, 100142, China. 3. Department of Radiology, Key Laboratory of Obstetric and Gynecologic and Pediatric Diseases and Birth Defects of Ministry of Education, West China Second University Hospital, Sichuan University, 20# South Renmin Road, Chengdu, 610041, Sichuan, China. 4. Department of Radiology, West China Hospital, Sichuan University, 37# Guo Xue Xiang, Chengdu, 610041, Sichuan, China. yangzg666@163.com.
Abstract
BACKGROUND: Cardiac myxomas (CMs) and thrombi are associated with high morbidity and mortality. These two conditions need totally different treatments. However, they are difficult to distinguish using naked eye. In clinical, misdiagnoses occur now and then. This study aimed to compare the characteristics of CMs and cardiac thrombi and investigate the value of a radiomics signature in distinguishing CMs from cardiac thrombi, based on cardiovascular contrast-enhanced computed tomography (CECT) images. METHODS: A total of 109 patients who had CMs (n = 59) and cardiac thrombi (n = 50) were enrolled in this retrospective study from 2009 to 2019. First, the lesion characteristics of cardiovascular CECT images were documented and compared by two radiologists. Then all patients were randomly allotted to either a primary group or a validation group according to a 7:3 ratio. Univariate analysis and the least absolute shrinkage and selection operator were used to select robust features. The best radiomics signature was constructed and validated using multivariate logistic regression. An independent clinical model was created for comparison. RESULTS: The best radiomics signature was developed using eight selected radiomics. The classification accuracies of the radiomics signature were 90.8% and 90.9%, and the area under the receiver operating characteristic curves were 0.969 and 0.926 in the training and testing cohorts, respectively. Cardiovascular CECT images showed that the two diseases had significant differences in location, surface, Hydrothorax, pericardial effusion and heart enlargement. The naked eye findings were used to create the clinical model. All metrics of the radiomics signature were higher than those of clinical model. CONCLUSIONS: Compared with clinical model, the radiomics signature based on cardiovascular CECT performed better in differentiating CMs and thrombi, suggesting that it could help improving the diagnostic efficiency.
BACKGROUND:Cardiac myxomas (CMs) and thrombi are associated with high morbidity and mortality. These two conditions need totally different treatments. However, they are difficult to distinguish using naked eye. In clinical, misdiagnoses occur now and then. This study aimed to compare the characteristics of CMs and cardiac thrombi and investigate the value of a radiomics signature in distinguishing CMs from cardiac thrombi, based on cardiovascular contrast-enhanced computed tomography (CECT) images. METHODS: A total of 109 patients who had CMs (n = 59) and cardiac thrombi (n = 50) were enrolled in this retrospective study from 2009 to 2019. First, the lesion characteristics of cardiovascular CECT images were documented and compared by two radiologists. Then all patients were randomly allotted to either a primary group or a validation group according to a 7:3 ratio. Univariate analysis and the least absolute shrinkage and selection operator were used to select robust features. The best radiomics signature was constructed and validated using multivariate logistic regression. An independent clinical model was created for comparison. RESULTS: The best radiomics signature was developed using eight selected radiomics. The classification accuracies of the radiomics signature were 90.8% and 90.9%, and the area under the receiver operating characteristic curves were 0.969 and 0.926 in the training and testing cohorts, respectively. Cardiovascular CECT images showed that the two diseases had significant differences in location, surface, Hydrothorax, pericardial effusion and heart enlargement. The naked eye findings were used to create the clinical model. All metrics of the radiomics signature were higher than those of clinical model. CONCLUSIONS: Compared with clinical model, the radiomics signature based on cardiovascular CECT performed better in differentiating CMs and thrombi, suggesting that it could help improving the diagnostic efficiency.
Authors: Mary L Grebenc; Melissa L Rosado-de-Christenson; Curtis E Green; Allen P Burke; Jeffrey R Galvin Journal: Radiographics Date: 2002 May-Jun Impact factor: 5.333