Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2025 Aug 18;15(1):30248.
doi: 10.1038/s41598-025-16331-8.

Deep learning for survival prediction in triple-negative breast cancer: development and validation in real-world cohorts

Yiyue Xu  1 , Butuo Li  1 , Bing Zou  1 , Bingjie Fan  1 , Shijiang Wang  1 , Jinming Yu  1 , Taotao Dong  2 , Linlin Wang  3
Affiliations

Deep learning for survival prediction in triple-negative breast cancer: development and validation in real-world cohorts

Yiyue Xu et al. Sci Rep. .

Abstract

Triple-negative breast cancer (TNBC) is an aggressive and heterogeneous disease, highlighting the need for better patient stratification to guide treatment. We developed a deep learning-based survival model and an individualized prognosis system using data from 37,818 TNBC patients in the SEER database (split into training [65%], validation [17.5%], and test [17.5%] sets). The survival model, built using the pysurvival algorithm, achieved strong performance (C-index: 0.824 in validation set, 0.816 in test set), outperforming traditional methods (CPH: 0.781 and 0.785; RSH: 0.779 and 0.766). External validation on a real-world cohort confirmed its robustness (C-index: 0.758). Our individualized prognosis system also showed higher predictive accuracy than traditional AJCC-TNM staging (AUC 0.821 vs. 0.771). These tools improve TNBC prognosis assessment, enable better patient stratification, and provide clinicians with significant treatment recommendations.

Keywords: Deep learning; Prognosis model; Survival prediction system; Triple-negative breast cancer.

PubMed Disclaimer

Conflict of interest statement

Declarations. Ethics approval: This research was approved by the Institutional Review Board of Shandong Cancer Hospital (No. SDTHEC202300318). This study was in compliance with the Declaration of Helsinki. Formal written informed consent was not required for this retrospective study. Competing interests: The authors declare no competing interests.

Figures

Fig. 1
Fig. 1
Correlation matrix of included variables. The correlation coefficient indicates strength of association, with higher absolute values representing stronger correlations. Positive values denote positive relationships, while negative values indicate inverse relationships.
Fig. 2
Fig. 2
The IBS of the deep learning model. The Brier score quantifies the accuracy of probabilistic predictions by measuring the mean squared difference between predicted survival probabilities and observed outcomes. Scores range from 0 (perfect prediction) to 1 (complete prediction failure). (a) The IBS of the deep learning model in the training set was 0.09. (b) The IBS of the deep learning model in the validation set was 0.09. (c) The IBS of the deep learning model in the test set was 0.09. (d)The IBS of the deep learning model in the real-world set was 0.11.
Fig. 3
Fig. 3
The IBS of random survival forest (RSF) model was 0.13.
Fig. 4
Fig. 4
The new prognostic system based on deep learning. (a) Patients were stratified into six classes based on the predicted risk scores. (b) Kaplan-Meier curves of patients staged with the new staging system. (c) The receiver operating characteristic (ROC) curve of the new staging system.
Fig. 5
Fig. 5
The traditional AJCC-TNM staging system. (a) Kaplan-Meier curve of patients staged with the AJCC-TNM staging system. (b) The receiver operating characteristic (ROC) curve of the AJCC-TNM staging system.
See this image and copyright information in PMC

References

    1. Ferlay, J. et al. Cancer statistics for the year 2020: an overview. Int. J. Cancer. 10.1002/ijc.33588 (2021). - PubMed
    1. Cocco, S. et al. Biomarkers in triple-negative breast cancer: State-of-the-art and future perspectives. Int. J. Mol. Sci.2110.3390/ijms21134579 (2020). - PMC - PubMed
    1. Torre, L. A., Siegel, R. L., Ward, E. M. & Jemal, A. Global cancer incidence and mortality rates and trends–An update. Cancer Epidemiol. Biomark. Prev.25, 16–27. 10.1158/1055-9965.Epi-15-0578 (2016). - PubMed
    1. Venkitaraman, R. Triple-negative/basal-like breast cancer: clinical, pathologic and molecular features. Expert Rev. Anticancer Ther.10, 199–207. 10.1586/era.09.189 (2010). - PubMed
    1. Edge, S. B. & Compton, C. C. The American Joint Committee on Cancer: the 7th edition of the AJCC cancer staging manual and the future of TNM. Ann. Surg. Oncol.17, 1471–1474. 10.1245/s10434-010-0985-4 (2010). - PubMed

Publication types

LinkOut - more resources