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

The DOI matches its Crossref record; 59 of the 60 references it cites were found in a registry or an index.

Long-term cancer survival prediction using multimodal deep learning

Open access
Vale-Silva, Luís A., Rohr, Karl · 2021 · Scientific Reports

Published version of a preprint

Abstract

The age of precision medicine demands powerful computational techniques to handle high-dimensional patient data. We present MultiSurv, a multimodal deep learning method for long-term pan-cancer survival prediction. MultiSurv uses dedicated submodels to establish feature representations of clinical, imaging, and different high-dimensional omics data modalities. A data fusion layer aggregates the multimodal representations, and a prediction submodel generates conditional survival probabilities for follow-up time intervals spanning several decades. MultiSurv is the first non-linear and non-proportional survival prediction method that leverages multimodal data. In addition, MultiSurv can handle missing data, including single values and complete data modalities. MultiSurv was applied to data from 33 different cancer types and yields accurate pan-cancer patient survival curves. A quantitative comparison with previous methods showed that Multisurv achieves the best results according to different time-dependent metrics. We also generated visualizations of the learned multimodal representation of MultiSurv, which revealed insights on cancer characteristics and heterogeneity.

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 · 10.1038/s41598-021-92799-4

Cites

59 of 59 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 The MultiSurv model architecture. Input data are all from the NCI Genomic Data Commons database, including up to six different data modalities. Each data modality is handled by a dedicated DL submodel, trained to generate modality-specific feature representations. A data fusion layer combines the generated feature representation vectors into a single fused representation. A final neural network takes the fused feature representation and outputs a conditional survival probability for each of a set of pre-defined follow-up time intervals. Taking the cumulative product of the set of conditional survival probabilities produces the predicted survival curve. CNN convolutional neural network, FC fully-connected neural network.
Figure 1 The MultiSurv model architecture. Input data are all from the NCI Genomic Data Commons database, including up to six different data modalities. Each data modality is handled by a dedicated DL submodel, trained to generate modality-specific feature representations. A data fusion layer combines the generated feature representation vectors into a single fused representation. A final neural network takes the fused feature representation and outputs a conditional survival probability for each of a set of pre-defined follow-up time intervals. Taking the cumulative product of the set of conditional survival probabilities produces the predicted survival curve. CNN convolutional neural network, FC fully-connected neural network.CC BY · PubMed Central
Figure 2 MultiSurv predictions allow construction of accurate survival curves. MultiSurv outputs patient survival predictions for the defined discrete-time follow up intervals. These can then be averaged to obtain group-wide survival predictions. (a) Survival curves constructed using Multisurv predictions for each patient in the test dataset diagnosed with one of four selected cancer types. One example patient is highlighted for each cancer type and the corresponding last follow up time point is annotated (as “Last follow up” if the patient is censored or “Death” if it corresponds to patient death). Highlighted patient codes are TCGA-HI-7169 for PRAD, TCGA-B0-5691 for KIRC, TCGA-29-1762 for OV, and TCGA-19-1390 for GBM. (b) Survival curves for the four example cancer types in (a) compared with Kaplan–Meier estimator outputs. (c) Survival curves for all patients in the test dataset compared with the Kaplan–Meier estimator output. (d) MultiSurv predictions allow accurate stratification of patient risk groups. Patients were split into low and high-risk groups according to MultiSurv’s first output interval risk prediction using the median value across all patients as the threshold. The two resulting groups have significantly different Kaplan–Meier estimates (log-rank test). The plot shows MultiSurv prediction averages overlayed on the Kaplan–Meier estimators. PRAD prostate adenocarcinoma, KIRC kidney renal clear cell carcinoma, OV ovarian serous cystadenocarcinoma, GBM glioblastoma multiforme.
Figure 2 MultiSurv predictions allow construction of accurate survival curves. MultiSurv outputs patient survival predictions for the defined discrete-time follow up intervals. These can then be averaged to obtain group-wide survival predictions. (a) Survival curves constructed using Multisurv predictions for each patient in the test dataset diagnosed with one of four selected cancer types. One example patient is highlighted for each cancer type and the corresponding last follow up time point is annotated (as “Last follow up” if the patient is censored or “Death” if it corresponds to patient death). Highlighted patient codes are TCGA-HI-7169 for PRAD, TCGA-B0-5691 for KIRC, TCGA-29-1762 for OV, and TCGA-19-1390 for GBM. (b) Survival curves for the four example cancer types in (a) compared with Kaplan–Meier estimator outputs. (c) Survival curves for all patients in the test dataset compared with the Kaplan–Meier estimator output. (d) MultiSurv predictions allow accurate stratification of patient risk groups. Patients were split into low and high-risk groups according to MultiSurv’s first output interval risk prediction using the median value across all patients as the threshold. The two resulting groups have significantly different Kaplan–Meier estimates (log-rank test). The plot shows MultiSurv prediction averages overlayed on the Kaplan–Meier estimators. PRAD prostate adenocarcinoma, KIRC kidney renal clear cell carcinoma, OV ovarian serous cystadenocarcinoma, GBM glioblastoma multiforme.CC BY · PubMed Central
Figure 3 Visualization of feature representations learned by MultiSurv. We collected the internal fused feature representation vector of the MultiSurv model trained on clinical and mRNA data and embedded it into a two-dimensional space using t-SNE. (a) Embedded feature representations for each patient in the test dataset. Patients diagnosed with each of four selected cancer types are highlighted. Within each of the highlighted cancer types, visually selected outlier patients are annotated. All patient survival curves, highlighting the selected outliers, are displayed for (b) PRAD, (c) KIRC, (d) GBM, and (e) OV. PRAD prostate adenocarcinoma, KIRC kidney renal clear cell carcinoma, OV ovarian serous cystadenocarcinoma, GBM glioblastoma multiforme.
Figure 3 Visualization of feature representations learned by MultiSurv. We collected the internal fused feature representation vector of the MultiSurv model trained on clinical and mRNA data and embedded it into a two-dimensional space using t-SNE. (a) Embedded feature representations for each patient in the test dataset. Patients diagnosed with each of four selected cancer types are highlighted. Within each of the highlighted cancer types, visually selected outlier patients are annotated. All patient survival curves, highlighting the selected outliers, are displayed for (b) PRAD, (c) KIRC, (d) GBM, and (e) OV. PRAD prostate adenocarcinoma, KIRC kidney renal clear cell carcinoma, OV ovarian serous cystadenocarcinoma, GBM glioblastoma multiforme.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 60 references

Research works printed but not linked: 1

  1. Oliphant, T. E. A Guide to NumPy Vol. 1 (Trelgol Publishing, 2006).

Cited by

See all 268

Metadata

Authors
Vale-Silva, L. A., & Rohr, K.
Affiliations
Heidelberg University (DE)
Corresponding author
Vale-Silva, Luís A.
Venue
Scientific Reports
Published
2021-06-29
Volume
11
Issue
1
Pages
13505-13505
ISSN
2045-2322 (ISSN-L)
OpenAlex
W3176016422
Cited by
268
References
60
Citations by year
2026: 47 · 2025: 78 · 2024: 67 · 2023: 43 · 2022: 26 · 2021: 7
FWCI
12.63
Type
journal article
Language
en
Open access
Licence
Locations
  • Scientific Reports (publishedVersion, cc-by)
  • medRxiv (acceptedVersion, cc-by-nc)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • Chica and Heinz Schaller Foundation (Personal fellowship to L.A.V.S.)
  • Ruprecht-Karls-Universität Heidelberg
Topics
  • Physical Sciences › Computer Science › Artificial Intelligence › Machine Learning in Healthcare
  • Health Sciences › Medicine › Radiology, Nuclear Medicine and Imaging › Radiomics and Machine Learning in Medical Imaging
  • Physical Sciences › Computer Science › Artificial Intelligence › AI in cancer detection
Keywords
cancer survival prediction, pan-cancer, multimodal deep learning, non-proportional hazards, conditional survival probability, data fusion, missing data, tumor heterogeneity
MeSH terms
Cancer Survivors, Deep Learning, Humans, Models, Biological, Neoplasms, Predictive Value of Tests, Survival Rate, Databases, Factual

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: Long-term cancer survival prediction using multimodal deep learning. https://app.cochat.ai/papers/10.1038/s41598-021-92799-4
The paper
Vale-Silva, L. A., & Rohr, K. (2021). Long-term cancer survival prediction using multimodal deep learning. Scientific Reports. https://doi.org/10.1038/s41598-021-92799-4

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.