Skip to Main Content

An official website of the United States government

Convolutional Survival Machines: Deep Convolutional Networks for Survival Analysis with Time-Varying risks.

Principal Investigator
Name
Artur Dubrawski
Degrees
M.Sc., M.Eng., Ph.D
Institution
Carnegie Mellon University
Position Title
Research Professor
Email
About this CDAS Project
Study
NLST (Learn more about this study)
Project ID
NLST-738
Initial CDAS Request Approval
Dec 21, 2020
Title
Convolutional Survival Machines: Deep Convolutional Networks for Survival Analysis with Time-Varying risks.
Summary
The use of Deep Learning based methods to assess patient risk and survival (Zhu, 2016; Zhu, 2017; Li, 2019; Mobadersany, 2018; Lu, 2019) from imaging modalities such as chest radiographs is gaining interest in the healthcare community. Current state of the art approaches rely on convolutional neural networks with Proportional Hazards assumption (Katzman, 2018) for model inference and learning. However, the proportional hazards assumption is restrictive and in many practical scenarios patients suffer from time varying risks that evolve over time. In this project, we will first study the time varying nature of mortality hazard rates of participants in 2 randomized clinical trials: the Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial (n = 10464) and National Lung Screening Trial (n = 5493) (Lu, 2019).

We will further propose extensions to our previous published research (Nagpal, 2020) involving deep learning methods to handle time varying risks that allow end to end learning on image data such as radiographs.
Aims

- Empirically Demonstrate and quantify scenarios where the proportional hazards assumptions are violated in the PLCO (Andriole, 2012) and NLST (National Lung Screening Trial Research Team, 2011) datasets due to the presence of time varying hazard rates.

- Propose and end to end learning framework to handle such scenarios using the parametric Deep Survival Machines model, using representations learned with convolutional neural networks,

- Add extensions to our existing open-source toolkit for deep learning based survival analysis, Deep Survival Machines (github.com/autonlab/DeepSurvivalMachines) to handle complex modalities like image data such as radiographs. We hope these extensions would allow the academic community to experiment and build risk stratification and survival models in other domains with complex data.

References:

Andriole, G. L., Crawford, E. D., Grubb III, R. L., Buys, S. S., Chia, D., Church, T. R., ... & Weissfeld, J. L. (2012). Prostate cancer screening in the randomized Prostate, Lung, Colorectal, and Ovarian Cancer Screening Trial: mortality results after 13 years of follow-up. Journal of the National Cancer Institute, 104(2), 125-132.

Katzman, J. L., Shaham, U., Cloninger, A., Bates, J., Jiang, T., & Kluger, Y. (2018). DeepSurv: Personalized treatment recommender system using a Cox proportional hazards deep neural network. BMC Medical Research Methodology, 18(1). doi:10.1186/s12874-018-0482-1

Li, H., Boimel, P., Janopaul-Naylor, J., Zhong, H., Xiao, Y., Ben-Josef, E., & Fan, Y. (2019). Deep Convolutional Neural Networks For Imaging Data Based Survival Analysis Of Rectal Cancer. 2019 IEEE 16th International Symposium on Biomedical Imaging (ISBI 2019). doi:10.1109/isbi.2019.8759301

Lu, M. T., Ivanov, A., Mayrhofer, T., Hosny, A., Aerts, H. J., & Hoffmann, U. (2019). Deep Learning to Assess Long-term Mortality From Chest Radiographs. JAMA Network Open, 2(7). doi:10.1001/jamanetworkopen.2019.7416

Mobadersany, P., Yousefi, S., Amgad, M., Gutman, D. A., Barnholtz-Sloan, J. S., Vega, J. E., . . . Cooper, L. A. (2017). Predicting cancer outcomes from histology and genomics using convolutional networks. doi:10.1101/198010

Nagpal, C., Li, X., & Dubrawski, A. (2020). Deep Survival Machines: Fully Parametric Survival Regression and Representation Learning for Censored Data with Competing Risks. arXiv preprint arXiv:2003.01176.

National Lung Screening Trial Research Team. (2011). The national lung screening trial: overview and study design. Radiology, 258(1), 243-253.

Zhu, X., Yao, J., & Huang, J. (2016). Deep convolutional neural network for survival analysis with pathological images. 2016 IEEE International Conference on Bioinformatics and Biomedicine (BIBM). doi:10.1109/bibm.2016.7822579

Zhu, X., Yao, J., Zhu, F., & Huang, J. (2017). WSISA: Making Survival Prediction from Whole Slide Histopathological Images. 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). doi:10.1109/cvpr.2017.725

Collaborators

Artur Dubrawski, Professor, Carnegie Mellon University
Chirag Nagpal, PhD Student, Carnegie Mellon University
Chufan Gao, MS Student, Carnegie Mellon University