Skip to main content
Tessa NewsLink
Paste a health news link, or browse
Latest checks

Source study found

Story checked

You can detect biological age and a person's chance of surviving cancer 'with just one photo' - Yahoo News Singapore (opens in a new tab)

sg.news.yahoo.com · 2024-01-01

Short answerEvidenceSource

Short answer

Not supported

Not supported.

One key claim is not backed by the study. 4 other points were not covered by the paper.

  • 2 overstated
  • 1 not supported
  • 4 not covered

Checked against the study summary. The full text wasn't available, so some details couldn't be settled either way.

Share this check

Follow the evidence trail
1

The story

You can detect biological age and a person's chance of surviving cancer 'with just one photo' - Yahoo News Singapore

sg.news.yahoo.com · 2024-01-01

The story’s checkable claims.

Read the original story (opens in a new tab)
2

NewsLink checks it

Not supported

Three claims go beyond the study. Two overstate it and one isn't supported at all. Four claims the study doesn't address.

  • 2 overstated
  • 1 not supported
  • 4 not covered
Open claim evidence
3
Then inspect each claim

Evidence layer

Claim by claim

Each claim gets a verdict. Expand it to see the evidence directly below.

7 claims in this story

Showing all 7 claimsChoose a verdict to focus the list.

Claim 1 of 7Not supportedAn AI tool developed by experts from Mass General Brigham in Massachusetts was trained on more than 40 million pictures to determine key indicators of a person's health.View evidenceHide evidence

As statedmore than 40 million pictures

Why this verdict

The abstract profile says FaceAge was trained/initially validated on 58,851 presumed healthy individuals aged 60 or older, including 56,304 from IMDb-Wiki for training and 2,547 from UTKFace for initial validation. It does not support the story’s claim that the tool was trained on more than 40 million pictures, and the broad phrase “determine key indicators of a person's health” is not supported by the abstract’s training description.

1

Study evidence

FaceAge was trained on data from 58,851 presumed healthy individuals aged 60 years or older: 56,304 individuals from the IMDb-Wiki dataset (training) and 2,547 from the UTKFace dataset (initial validation).

“FaceAge was trained on data from 58 851 presumed healthy individuals aged 60 years or older: 56 304 individuals from the IMDb-Wiki dataset (training) and 2547 from the UTKFace dataset (initial validation).”
Claim 2 of 7OverstatedA person's biological age and their likelihood of surviving cancer can both be detected from a single photograph of their face, according to researchers.View evidenceHide evidence

Why this verdict

The abstract supports that FaceAge estimates biological/visual age from face photographs and that higher FaceAge is associated with worse overall survival in cancer cohorts, with added prognostic value in a palliative cohort. But the story’s lead wording says a person's biological age and cancer-survival likelihood can be “detected” from a single photo, which is stronger and broader than the abstract-level evidence: the paper reports model-based estimates and observational prognostic associations, not definitive individual detection for any person.

1

Study evidence

FaceAge was trained on data from 58,851 presumed healthy individuals aged 60 years or older: 56,304 individuals from the IMDb-Wiki dataset (training) and 2,547 from the UTKFace dataset (initial validation).

“FaceAge was trained on data from 58 851 presumed healthy individuals aged 60 years or older: 56 304 individuals from the IMDb-Wiki dataset (training) and 2547 from the UTKFace dataset (initial validation).”
2

Study evidence

Higher FaceAge (appearing older) was independently associated with worse overall survival after adjustment: per-decade HR 1.151 (p=0.013) in a pan-cancer cohort (n=4,906).HR 1.151 per decade

“Clinical utility was evaluated on data from 6196 patients with cancer diagnoses from two institutions in the Netherlands and the USA: the MAASTRO, Harvard Thoracic, and Harvard Palliative cohorts”
Claim 3 of 7OverstatedThe authors said the cancer estimates were checked against real patient data to verify their accuracy.View evidenceHide evidence

Why this verdict

The abstract supports that FaceAge was evaluated against real cancer-patient data in 6,196 patients and that survival associations and prediction-model performance were assessed. However, saying the estimates were checked “to verify their accuracy” is stronger than the abstract evidence, which reports observational cohort associations and AUC improvement in a palliative cohort while noting the need for further validation.

1

Study evidence

Higher FaceAge (appearing older) was independently associated with worse overall survival after adjustment: per-decade HR 1.151 (p=0.013) in a pan-cancer cohort (n=4,906).HR 1.151 per decade

“Clinical utility was evaluated on data from 6196 patients with cancer diagnoses from two institutions in the Netherlands and the USA: the MAASTRO, Harvard Thoracic, and Harvard Palliative cohorts”
2

Study evidence

In patients with incurable/palliative cancer, incorporation of FaceAge into prediction models improved physicians' survival-prediction discrimination from AUC 0.74 (95% CI 0.70–0.78) to 0.80 (95% CI 0.76–0.83); p<0.0001 (palliative cohort n=717).AUC increase 0.06 (from 0.74 to 0.80); p<0.0001

“We also assessed the performance of FaceAge in patients with metastatic cancer receiving palliative treatment at the end of life by incorporating FaceAge into clinical prediction models.”
Claim 4 of 7Not coveredHeavy alcohol use, drug use and smoking were associated with looking biologically older, with smoking showing the strongest pattern.View evidenceHide evidence

Why this verdict

The abstract-level paper profile contains no evidence about heavy alcohol use, drug use, smoking, or smoking being the strongest lifestyle correlate of looking biologically older. This may require full-text evidence, but it is not verifiable from the supplied abstract-depth profile.

Claim 5 of 7Not coveredHugo Aerts said the work advances facial health recognition AI and is, to their knowledge, the first demonstration that an AI model can directly predict cancer survival from a simple photograph.View evidenceHide evidence

Why this verdict

The abstract supports the general basis for a survival-prediction claim: FaceAge was associated with overall survival and improved physician survival-prediction AUC in a palliative cohort. But the exact priority claim—“to our knowledge, the first demonstration”—and the quoted wording are not verifiable from the supplied abstract-depth profile. The abstract evidence also remains observational and calls for further validation.

1

Study evidence

Higher FaceAge (appearing older) was independently associated with worse overall survival after adjustment: per-decade HR 1.151 (p=0.013) in a pan-cancer cohort (n=4,906).HR 1.151 per decade

“Clinical utility was evaluated on data from 6196 patients with cancer diagnoses from two institutions in the Netherlands and the USA: the MAASTRO, Harvard Thoracic, and Harvard Palliative cohorts”
2

Study evidence

In patients with incurable/palliative cancer, incorporation of FaceAge into prediction models improved physicians' survival-prediction discrimination from AUC 0.74 (95% CI 0.70–0.78) to 0.80 (95% CI 0.76–0.83); p<0.0001 (palliative cohort n=717).AUC increase 0.06 (from 0.74 to 0.80); p<0.0001

“We also assessed the performance of FaceAge in patients with metastatic cancer receiving palliative treatment at the end of life by incorporating FaceAge into clinical prediction models.”
Claim 6 of 7Not coveredRaymond Mak said the findings were consistent across common real-world photo variations such as pose, lighting, facial expression and makeup.View evidenceHide evidence

Why this verdict

The supplied abstract profile does not report robustness or consistency testing across pose, lighting, facial expression, makeup, or other real-world photo variations. At abstract depth this claim cannot be verified.

1

Study evidence

FaceAge was trained on data from 58,851 presumed healthy individuals aged 60 years or older: 56,304 individuals from the IMDb-Wiki dataset (training) and 2,547 from the UTKFace dataset (initial validation).

“FaceAge was trained on data from 58 851 presumed healthy individuals aged 60 years or older: 56 304 individuals from the IMDb-Wiki dataset (training) and 2547 from the UTKFace dataset (initial validation).”
Claim 7 of 7Not coveredThe researchers said the goal is to make face photograph-based health assessment more quantitative, consistent and scalable so it can support personalised treatment decisions rather than replace physician judgment.View evidenceHide evidence

Why this verdict

The abstract profile is broadly consistent with making visual ageing more objective and quantitative and with using FaceAge to support prognostication, especially in palliative cancer care. But the specific quoted goal—quantitative, consistent, scalable assessment supporting personalised treatment decisions rather than replacing physician judgment—is not stated in the supplied abstract-depth profile, so the full claim is not verifiable at this depth.

1

Study evidence

FaceAge was trained on data from 58,851 presumed healthy individuals aged 60 years or older: 56,304 individuals from the IMDb-Wiki dataset (training) and 2,547 from the UTKFace dataset (initial validation).

“FaceAge was trained on data from 58 851 presumed healthy individuals aged 60 years or older: 56 304 individuals from the IMDb-Wiki dataset (training) and 2547 from the UTKFace dataset (initial validation).”
2

Study evidence

In patients with incurable/palliative cancer, incorporation of FaceAge into prediction models improved physicians' survival-prediction discrimination from AUC 0.74 (95% CI 0.70–0.78) to 0.80 (95% CI 0.76–0.83); p<0.0001 (palliative cohort n=717).AUC increase 0.06 (from 0.74 to 0.80); p<0.0001

“We also assessed the performance of FaceAge in patients with metastatic cancer receiving palliative treatment at the end of life by incorporating FaceAge into clinical prediction models.”
Then look for missing context

Context layer

What the story left out

Important study details the story did not include.

  • FaceAge model development used 58,851 presumed healthy individuals aged 60 years or older, with 56,304 IMDb-Wiki individuals for training and 2,547 UTKFace individuals for initial validation.

    The story reflects that an AI face-photo model was developed, but it reports a materially different training scale of more than 40 million pictures and does not mention the abstract’s age restriction or presumed-healthy training/validation population.

    From model development and initial validation

  • In patients with incurable or palliative cancer, adding FaceAge improved physician survival-prediction discrimination from AUC 0.74 to 0.80.

    The story says the tool is intended to support personalised treatment decisions rather than replace physicians, but it does not reflect the paper’s specific incremental-prediction result, the palliative/incurable cancer setting, or the AUC magnitude.

    From observational prognostic model augmentation

  • Patients with cancer appeared older than a non-cancer reference cohort by a mean 4.79 years.

    This cross-sectional cancer-versus-reference finding is a material result in the abstract profile, but it is not captured in the presented story claims.

    From observational cross-sectional comparison

  • The abstract reports a gene-based analysis in which FaceAge, but not chronological age, was significantly associated with senescence-related molecular mechanisms.

    The story does not mention the molecular ageing or senescence-gene analysis, which is a secondary contribution in the paper profile.

    From gene-based/senescence association analysis

  • The abstract notes a need for further research and validation in larger cohorts before broader clinical application.

    The story’s caveats mention real-patient checking, photo-variation consistency, and physician-support framing, but it does not reflect the paper’s stated need for further validation in larger cohorts.

    From model development and initial validation; Observational cohort prognostic analysis; observational prognostic model augme

  • Generalizability is limited at abstract depth: training/initial validation involved presumed healthy people aged 60 or older, cancer analyses used specific institutional cohorts, and representativeness, demographics, fairness, and image-acquisition details are not described.

    The story’s broad wording about “a person” and its emphasis on robustness across photo variations do not reflect these generalizability limitations, which are material to interpreting the results.

    From model development and initial validation; Observational cohort prognostic analysis; observational cross-sectional compar

1 thing the story did carry across
  • The paper’s cancer-survival evidence is observational: higher FaceAge was associated with worse overall survival after covariate adjustment across pan-cancer, thoracic, and palliative cohorts.
Then read the study layer

Study layer

Study at a glance

Scan the study first. Expand only the parts you want to inspect.

Pieces of work

5

Evidence read

study summary

Lead result

secondary data

1Lead resultsecondary dataDevelop and validate FaceAge, a deep learning system that estimates biological age from face photographs using large datasets of presumed healthy older adults.model development and initial validationExpand

In plain English

Development and initial validation of FaceAge, a supervised deep-learning system to estimate biological (visual) age from face photographs. The model was trained on a large set of presumed-healthy older adults (total n=58,851; IMDb-Wiki n=56,304 for training) and initially validated on UTKFace (n=2,547). The abstract describes the training/validation datasets and the modelling aim but does not provide model architecture, training procedures, or performance metrics for the initial validation within the abstract.

Key findings

  • FaceAge was trained on data from 58,851 presumed healthy individuals aged 60 years or older: 56,304 individuals from the IMDb-Wiki dataset (training) and 2,547 from the UTKFace dataset (initial validation).
  • The abstract reports that the system estimates biological age from face photographs using supervised deep learning.
“FaceAge was trained on data from 58 851 presumed healthy individuals aged 60 years or older: 56 304 individuals from the IMDb-Wiki dataset (training) and 2547 from the UTKFace dataset (initial validation).”
What this piece can’t prove
  • Age range restricted to individuals aged ≥60 years in the cited training/validation datasets—limits applicability to younger populations.
  • Initial validation reported only on UTKFace (n=2,547); external generalizability beyond these datasets is not established in this unit.

2 further details could not be confirmed from the summary.

2human in vivoEvaluate clinical/prognostic utility of FaceAge in multiple cancer patient cohorts, including survival association and comparison to a non-cancer reference cohort.Observational cohort prognostic analysisExpand

In plain English

The authors evaluated the prognostic utility of FaceAge (a deep-learning estimate of biological age from face photographs) in 6,196 patients with cancer from three cohorts (MAASTRO, Harvard Thoracic, Harvard Palliative). Time-to-event analyses (Kaplan–Meier and multivariable Cox models) showed that higher FaceAge (looking older) was independently associated with worse overall survival across a pan-cancer cohort and in thoracic and palliative subcohorts. Patients with cancer had higher FaceAge than a non-cancer reference cohort. Incorporating FaceAge improved physicians' survival prediction performance in patients receiving palliative treatment.

Key findings

  • Higher FaceAge (appearing older) was independently associated with worse overall survival after adjustment: per-decade HR 1.151 (p=0.013) in a pan-cancer cohort (n=4,906).HR 1.151 per decade
  • Similar adjusted associations were observed in subgroup cohorts: thoracic cohort per-decade HR 1.148 (p=0.011; n=573) and palliative cohort per-decade HR 1.117 (p=0.021; n=717).HR 1.148 (thoracic); HR 1.117 (palliative)
“Clinical utility was evaluated on data from 6196 patients with cancer diagnoses from two institutions in the Netherlands and the USA: the MAASTRO, Harvard Thoracic, and Harvard Palliative cohorts”
What this piece can’t prove
  • The authors note the need for further validation in larger cohorts and to assess generalisability beyond the studied cancer populations.

3 further details could not be confirmed from the summary.

3human in vivoEvaluate clinical/prognostic utility of FaceAge in multiple cancer patient cohorts, including survival association and comparison to a non-cancer reference cohort.observational cross-sectional comparisonExpand

In plain English

Cross-sectional comparison of FaceAge-derived age gap (FaceAge minus chronological age) between pooled cancer cohorts (n=6196) and a non-cancer reference cohort (n=535) found that patients with cancer appeared older on average by 4.79 years (p<0.0001).

Key findings

  • Patients with cancer appeared older than their chronological age compared with a non-cancer reference cohort.Mean increase 4.79 years vs non-cancer reference, p<0.0001
“FaceAge estimates in these cancer cohorts were compared with a non-cancerous reference cohort of 535 individuals.”
What this piece can’t prove
  • Cross-sectional design of this comparison limits causal inference.
  • Authors state the need for validation in larger cohorts, indicating the finding is preliminary at the reported scale.

2 further details could not be confirmed from the summary.

4human in vivoAssess whether FaceAge improves end-of-life prognostication by incorporating it into clinical prediction models (including physician predictions) in metastatic/palliative cancer cohorts.observational prognostic model augmentationExpand

In plain English

In a palliative/metastatic cancer setting, the authors evaluated whether adding FaceAge (a face-photograph–based biological age estimate) to clinical prediction models improves end-of-life survival prognostication, including comparison with physician predictions. In the palliative cohort (n=717) incorporation of FaceAge increased physician survival-prediction discrimination from AUC 0.74 (95% CI 0.70–0.78) to 0.80 (95% CI 0.76–0.83); p<0.0001. The abstract reports this as evidence that FaceAge can enhance prognostic models for patients receiving palliative treatments, but calls for further validation in larger cohorts.

Key findings

  • In patients with incurable/palliative cancer, incorporation of FaceAge into prediction models improved physicians' survival-prediction discrimination from AUC 0.74 (95% CI 0.70–0.78) to 0.80 (95% CI 0.76–0.83); p<0.0001 (palliative cohort n=717).AUC increase 0.06 (from 0.74 to 0.80); p<0.0001
“We also assessed the performance of FaceAge in patients with metastatic cancer receiving palliative treatment at the end of life by incorporating FaceAge into clinical prediction models.”
What this piece can’t prove
  • Authors note need for further validation in larger cohorts.
  • Unclear from abstract whether AUC improvement translates to patient-centered clinical benefit or changes in decision-making outcomes.

1 further detail could not be confirmed from the summary.

5in silicoEvaluate whether FaceAge is associated with molecular ageing mechanisms via gene-based analysis of associations with senescence genes.gene-based/senescence association analysisExpand

In plain English

The authors report a gene-based, in-silico analysis assessing whether FaceAge estimates associate with senescence-related molecular mechanisms. In the abstract they state that FaceAge was significantly associated with molecular mechanisms of senescence through gene analysis, whereas chronological age was not. The abstract provides no further methodological or quantitative detail about the gene-based testing.

Key findings

  • FaceAge was significantly associated with molecular mechanisms of senescence through gene analysis, whereas chronological age was not.
“To evaluate whether FaceAge has the potential to be a biomarker for molecular ageing, we performed a gene-based analysis to assess its association with senescence genes.”
What this piece can’t prove
  • Limited methodological and quantitative detail in abstract for the gene-based analysis.
  • Unknown genomic dataset, sample size, gene sets, and statistical procedures; cannot assess robustness, correction for multiple testing, or potential confounding.

1 further detail could not be confirmed from the summary.

Finally, the search trail

Method layer

NewsLink found the paper. Tessa takes you deeper.

NewsLink checks the story. Tessa is where you inspect the paper, authors, evidence, and research context.

Papers considered

The selected paper, plus nearby candidates.

PubMed, Europe PMC, Crossref · 15 candidate papers

Candidate

FaceAge+: training-free synthetic augmentation for deep learning-based biological age estimation from facial images

Scientific Reports · 2026 · Crossref

Source (opens in a new tab)
Candidate

Associations Between G8 Geriatric Screening Score, Charlson Comorbidity Index, AI-Based Age Phenotype, and Overall Survival in Older Adults With Stage I-II Non-Small Cell Lung Cancer.

International Journal of Radiation Oncology, Biology, Physics · 2025 · PubMed, Europe PMC

Tessa (opens in a new tab)Source (opens in a new tab)

And 9 more candidates considered.