Abstract 130P
Background
There are no satisfying ways to distinguish high- from low-risk patients with early-stage breast cancer. We aimed to develop a MRI radiomic-based signature (MRS) for predicting prognosis and discriminating of high-risk relapse patients based on deep learning approach.
Methods
A total of 452 patients with early-stage breast cancer from Sun Yat-sen Memorial Hospital (SYSMH) were randomly assigned (3:1) into the training and the internal validation cohorts, and the training cohort was then used to optimize the model parameters. The performance of the siganature for disease-free survival (DFS) prediction was validated in SYSMH internal validation cohort (n=113) and two independent external validation cohorts: Sun Yat-sen University Cancer Center (SYSUCC) (n=325), and Tungwah Hospital of Sun Yat-Sen University and Shunde Hospital of Southern Medical University (DHSDH) (n=163). The concordance index (C-index) and receiver operating characteristic (ROC) curves were used to evaluate the performance of the signature.
Results
In the SYSMH internal validation cohort, the MRS based on T1+C and T2WI sequences showed a higher prediction quality, with C-index of 0.935, 1-, 2-, 3-year AUCs of 0.981, 0.962, and 0.996 compared with the signature based on T1+C or T2WI single sequence. The prognostic value of MRS was well validated that it could discriminate high- from low-risk patients in the SYSUCC external validation (C-index: 0.828; 1-, 2-, 3-year AUCs: 0.944, 0.926, and 0.806) and DHSDH external validation (C-index: 0.870; 1-, 2-, 3-year AUCs: 0.921, 0.925, and 0.960) cohorts. MRS was also found to be significantly associated with DFS in the SYSMH internal validation cohort (HR 0.009, 95% CI 0.001-0.077, P < 0.001), the SYSUCC external validation cohort (HR 0.129, 95% CI 0.043-0.384, P < 0.001), and the DHSDH external validation cohort (HR 0.033, 95% CI 0.010-0.111, P < 0.001).
Conclusions
Our study offers a noninvasive imaging biomarker for DFS prediction and the MRS shows the potential to be served as a convenient tool for identifying high-risk relapse individuals in patients with early-stage invasive breast cancer.
Clinical trial identification
Editorial acknowledgement
Legal entity responsible for the study
The authors.
Funding
This study was supported by grant 2020ZX09201021 from the National Science and Technology Major Project, grant YXRGZN201902 from the Medical Artificial Intelligence Project of Sun Yat-Sen Memorial Hospital, grants 81572596, 81972471, and 82073408 from the National Natural Science Foundation of China, grant 2017A030313828 from the Natural Science Foundation of Guangdong Province, grant 201704020131 from the Guangzhou Science and Technology Major Program, grant 2017B030314026 from the Guangdong Science and Technology Department, grant 2018007 from the Sun Yat-Sen University Clinical Research 5010 Program, grant SYS-C-201801 from the Sun Yat-Sen Clinical Research Cultivating Program, grant A2020558 from the Guangdong Medical Science and Technology Program.
Disclosure
All authors have declared no conflicts of interest.