1
|
Zhou X, Gao F, Duan S, Zhang L, Liu Y, Zhou J, Bai G, Tao W. Radiomic features of Pk-DCE MRI parameters based on the extensive Tofts model in application of breast cancer. Phys Eng Sci Med 2020; 43:517-524. [PMID: 32524436 DOI: 10.1007/s13246-020-00852-9] [Citation(s) in RCA: 2] [Impact Index Per Article: 0.5] [Reference Citation Analysis] [Abstract] [Key Words] [Track Full Text] [Journal Information] [Subscribe] [Scholar Register] [Received: 03/12/2019] [Accepted: 02/11/2020] [Indexed: 01/03/2023]
Abstract
To explore radiomic features of pharmacokinetic dynamic contrast-enhanced (Pk-DCE) MRI on the extensive Tofts model to diagnose breast cancer and predict molecular phenotype. Breast lesions enrolled must undergo Pk-DCE MRI before treatment or puncture, and be identified as primary lesions by pathology. Ktrans, Kep, Ve and Vp were generated on the extensive Tofts model. Radiomic features (histogram, geometry and texture features) were extracted from parametric maps and selected by LASSO. The subjects were divided into training and validation cohort with a ratio of 4:1 to construct model in diagnosis of breast cancer. Feature analysis was made to predict the molecular phenotype. Area under curve (AUC), sensitivity, specificity and accuracy were used to evaluate radiomic features. DeLong's test was performed to compare AUC values. 228 breast lesions met the criteria were used to discrimination and 126 malignant lesions were used to study molecular phenotypes. The number of training cohort and validation cohort were 182 and 46, respectively. The AUC of Ktrans, Kep, Ve, and Vp was 0.95, 0.93, 0.89, and 0.96, and their accuracy was 85%, 89%, 89%, 94% respectively in diagnosis of breast lesions, while their AUC was 0.71 to 0.77, 0.61 to 0.68, and 0.67 to 0.74 to predict ER/PR, Her-2, and Ki-67. There was no significant difference among parameters (P > 0.05). Radiomic features based on Pk-DCE MRI have an advantage to diagnose breast cancer and less ability to predict molecular phenotypes, which are beneficial to guide clinical treatment of breast lesions in some extent.
Collapse
Affiliation(s)
- Xiaoyu Zhou
- Research Center of Internet Things (Sensory Mine), China University of Mining and Technology, Xuzhou, People's Republic of China.,Faculty of Mechanical Electronic and Information Engineering, Jiangsu Vocational College of Finance and Economics, Huai'an, People's Republic of China
| | - Feng Gao
- Department of Medical Imaging, Jinling Hospital, Medical School of Nanjing University, Nanjing, People's Republic of China
| | - Shaofeng Duan
- GE Healthcare China, Shanghai, People's Republic of China
| | - Lianmei Zhang
- Department of Pathology, The Affiliated Huai'an No. 1 People's Hospital of Nanjing Medical University, Huai'an, People's Republic of China
| | - Yan Liu
- Department of Radiology, The Affiliated Huai'an No. 1 People's Hospital of Nanjing Medical University, Huanghe Road No. 1, Huai'an, 223300, Jiangsu Province, People's Republic of China
| | - Junyi Zhou
- Department of Medical Imaging, Jiangsu University, Zhenjiang, People's Republic of China
| | - Genji Bai
- Department of Radiology, The Affiliated Huai'an No. 1 People's Hospital of Nanjing Medical University, Huanghe Road No. 1, Huai'an, 223300, Jiangsu Province, People's Republic of China.
| | - Weijing Tao
- Department of Nuclear Medicine, The Affiliated Huai'an No. 1 People's Hospital of Nanjing Medical University, Huanghe Road No.1, Huai'an, 223300, Jiangsu Province, People's Republic of China.
| |
Collapse
|