Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 36 of the 37 references it cites were found in a registry or an index, and 1 are not checked yet.

Deep learning-based survival prediction for multiple cancer types using histopathology images

Open access

Abstract

Providing prognostic information at the time of cancer diagnosis has important implications for treatment and monitoring. Although cancer staging, histopathological assessment, molecular features, and clinical variables can provide useful prognostic insights, improving risk stratification remains an active research area. We developed a deep learning system (DLS) to predict disease specific survival across 10 cancer types from The Cancer Genome Atlas (TCGA). We used a weakly-supervised approach without pixel-level annotations, and tested three different survival loss functions. The DLS was developed using 9,086 slides from 3,664 cases and evaluated using 3,009 slides from 1,216 cases. In multivariable Cox regression analysis of the combined cohort including all 10 cancers, the DLS was significantly associated with disease specific survival (hazard ratio of 1.58, 95% CI 1.28-1.70, p<0.0001) after adjusting for cancer type, stage, age, and sex. In a per-cancer adjusted subanalysis, the DLS remained a significant predictor of survival in 5 of 10 cancer types. Compared to a baseline model including stage, age, and sex, the c-index of the model demonstrated an absolute 3.7% improvement (95% CI 1.0-6.5) in the combined cohort. Additionally, our models stratified patients within individual cancer stages, particularly stage II (p = 0.025) and stage III (p<0.001). By developing and evaluating prognostic models across multiple cancer types, this work represents one of the most comprehensive studies exploring the direct prediction of clinical outcomes using deep learning and histopathology images. Our analysis demonstrates the potential for this approach to provide significant prognostic information in multiple cancer types, and even within specific pathologic stages. However, given the relatively small number of cases and observed clinical events for a deep learning task of this type, we observed wide confidence intervals for model performance, thus highlighting that future work will benefit from larger datasets assembled for the purposes for survival modeling.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1371/journal.pone.0233678

Cites

36 of 37 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig 1 Summary of the weakly supervised learning approach for directly predicting disease specific survival. For each case, cropped image patches were uniformly sampled from tissue-containing areas across all slides available for a given case. Next, image features were extracted for each patch by a convolutional neural network (CNN). These patch-level features were averaged (on a per-channel basis) and fed to a fully connected layer. Our custom loss function divided the follow-up period into four discrete bins depending on right-censorship time and outcome (Methods). As such, the model was designed to output a probability distribution over discretized survival times.
Fig 1 Summary of the weakly supervised learning approach for directly predicting disease specific survival. For each case, cropped image patches were uniformly sampled from tissue-containing areas across all slides available for a given case. Next, image features were extracted for each patch by a convolutional neural network (CNN). These patch-level features were averaged (on a per-channel basis) and fed to a fully connected layer. Our custom loss function divided the follow-up period into four discrete bins depending on right-censorship time and outcome (Methods). As such, the model was designed to output a probability distribution over discretized survival times.CC BY · PubMed Central
Fig 2 Kaplan Meier curves for DLS risk groups. To define low and high risk groups, cases were binned into risk quartiles using DLS risk scores. Binning was done within each cancer type to ensure that the distribution of cancer types within each risk group was the same. Different colors represent the different risk groups: green for the low risk (0th -25th percentile); yellow for medium risk (25th-75th percentile), and red for high risk (75th-100th percentile). P-values were calculated using the binary logrank test comparing the low and high risk groups. The Kaplan Meier curve for each individual cancer type is shown in S2 Fig.
Fig 2 Kaplan Meier curves for DLS risk groups. To define low and high risk groups, cases were binned into risk quartiles using DLS risk scores. Binning was done within each cancer type to ensure that the distribution of cancer types within each risk group was the same. Different colors represent the different risk groups: green for the low risk (0th -25th percentile); yellow for medium risk (25th-75th percentile), and red for high risk (75th-100th percentile). P-values were calculated using the binary logrank test comparing the low and high risk groups. The Kaplan Meier curve for each individual cancer type is shown in S2 Fig.CC BY · PubMed Central
Fig 3 Kaplan Meier curves for DLS risk groups within each cancer stage. To define low and high risk groups, cases were binned into risk quartiles using DLS risk scores. Binning was done within each stage and cancer type. This ensures that for each stage, the distribution of cancer types within each risk group was the same. P-values were calculated using the binary logrank test comparing the low and high risk groups. Unfortunately, there were insufficient cases and events to meaningfully perform this analysis stratified by cancer type in addition to stage.
Fig 3 Kaplan Meier curves for DLS risk groups within each cancer stage. To define low and high risk groups, cases were binned into risk quartiles using DLS risk scores. Binning was done within each stage and cancer type. This ensures that for each stage, the distribution of cancer types within each risk group was the same. P-values were calculated using the binary logrank test comparing the low and high risk groups. Unfortunately, there were insufficient cases and events to meaningfully perform this analysis stratified by cancer type in addition to stage.CC BY · PubMed Central
Fig 4 Visualization of image patches influencing survival prediction. (A) Example of WSI kidney renal clear cell carcinoma (KIRC) predicted to be high risk (left), with the DLS-predicted “risk heatmap” on the right; red patches correspond to “high-risk” and blue patches to “low-risk” patch-level predictions (Methods). (B) “Highest-risk” patches from cases predicted to be high-risk. (C) “Lowest-risk” patches from cases predicted to be low-risk. (D) “Lowest-risk” patches from cases predicted to be high-risk. For B, C, and D, patches in the same row are from the same case and each row represents a different case.
Fig 4 Visualization of image patches influencing survival prediction. (A) Example of WSI kidney renal clear cell carcinoma (KIRC) predicted to be high risk (left), with the DLS-predicted “risk heatmap” on the right; red patches correspond to “high-risk” and blue patches to “low-risk” patch-level predictions (Methods). (B) “Highest-risk” patches from cases predicted to be high-risk. (C) “Lowest-risk” patches from cases predicted to be low-risk. (D) “Lowest-risk” patches from cases predicted to be high-risk. For B, C, and D, patches in the same row are from the same case and each row represents a different case.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 36 references

Research works printed but not linked: 2

  1. Y Liu, 2018, Artificial Intelligence-Based Breast Cancer Nodal Metastasis Detection, Arch Pathol Lab Med
  2. T Tieleman, 2012, Lecture 6.5-rmsprop: Divide the gradient by a running average of its recent magnitude, COURSERA: Neural networks for machine learning, 4, 26

Cited by

See all 266

Metadata

Authors
Wulczyn, E., Steiner, D. F., Xu, Z., Sadhwani, A., Wang, H., Flament-Auvigne, I., Mermel, C. H., Chen, P.-H. C., Liu, Y., & Stumpe, M. C.
Affiliations
Google (United States) (US)
Corresponding author
Liu, Yun
Venue
PLOS ONE
Published
2020-06-17
Volume
15
Issue
6
Pages
e0233678
ISSN
1932-6203 (ISSN-L)
OpenAlex
W3036901136
Cited by
266
References
36
Citations by year
2026: 39 · 2025: 33 · 2024: 53 · 2023: 36 · 2022: 51 · 2021: 46 · 2020: 7
FWCI
13.75
Type
journal article
Language
en
Open access
Locations
  • PLoS ONE (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • arXiv (Cornell University) (submittedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • arXiv (Cornell University)
Funders
Google
Topics
  • Physical Sciences › Computer Science › Artificial Intelligence › AI in cancer detection
  • Health Sciences › Medicine › Radiology, Nuclear Medicine and Imaging › Radiomics and Machine Learning in Medical Imaging
  • Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning in Healthcare
Keywords
cancer survival prediction, histopathological images, disease-specific survival, prognostic model, weakly supervised learning, risk stratification, tumor staging, multiple cancer types, TCGA, deep learning
MeSH terms
Deep Learning, Adult, Age Factors, Feasibility Studies, Female, Humans, Image Processing, Computer-Assisted, Male, Middle Aged, Neoplasm Staging, Neoplasms, Prognosis, Risk Factors, Sex Factors, Survival Analysis, Risk Assessment, Datasets as Topic

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Deep learning-based survival prediction for multiple cancer types using histopathology images. https://app.cochat.ai/papers/10.1371/journal.pone.0233678
The paper
Wulczyn, E., Steiner, D. F., Xu, Z., Sadhwani, A., Wang, H., Flament-Auvigne, I., Mermel, C. H., Chen, P.-H. C., Liu, Y., & Stumpe, M. C. (2020). Deep learning-based survival prediction for multiple cancer types using histopathology images. PLOS ONE. https://doi.org/10.1371/journal.pone.0233678

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.