Can a Simple Photograph Help Predict Cancer Survival? The FaceAge Artificial Intelligence

Face Age: A Deep Learning System in Cancer Diagnosis

Published on May 8, 2025 in The Lancet Digital Health, this study shows that the deep learning system called FaceAge can predict biological age from facial photographs and improve survival predictions in cancer patients.

In clinical practice, the “first impression” of a patient may influence a doctor’s decisions, but it remains subjective and only provides a rough estimate of biological age. There was a need for a quantitative and standardized measure. FaceAge aims to fill this gap by generating a measurable “face age” from a single facial photo.

Summary: Compared with non-cancer individuals, cancer patients appear on average about 5 years older in facial photographs. Face age is independently associated with survival across multiple cancer cohorts. When AI is applied in palliative care, the accuracy of doctors’ survival predictions improves by about 6%. In other words, with AI support, prediction accuracy rises from 74% to 80%.

FaceAge generates an age prediction from a facial photo taken with a standard webcam/smartphone, adding an objective layer to the doctor’s subjective performance assessment. Patients whose face age is younger than their chronological age tend to have better outcomes after treatment.

🔹 Age difference: Compared with age-matched healthy individuals, cancer patients appeared on average 4.7 years older (P<0.0001).

🔹 Survival relationship (FaceAge ↑ → risk ↑):

• All cancer types combined: 15% higher mortality risk (HR 1.151; P=0.013)
• Thoracic cancers: 11% higher risk (HR 1.117; P=0.021)
• Palliative care patients: 11% higher risk (HR 1.117; P=0.021)

🔹 Clinical benefit: In palliative care, doctors’ 6-month survival prediction accuracy improved from 74% to 80%, representing a 6% absolute improvement (P<0.0001).

👉 In summary: Looking older than one’s actual age is linked with worse prognosis in cancer. By capturing this biological aging signal, the FaceAge algorithm significantly strengthens doctors’ ability to predict survival.

CDK6 (Cyclin-Dependent Kinase 6) is a critical enzyme regulating the cell cycle, playing a major role in cellular proliferation and aging mechanisms. The study highlighted the following findings:

• Negative correlation between FaceAge and CDK6 expression: Patients who appeared older than their chronological age had lower CDK6 levels.
• Chronological age (calendar age) showed no significant correlation with CDK6 after multiple statistical corrections.

What does this mean?

• Lower CDK6 levels may indicate reduced cellular regeneration and an accelerated biological aging process.
• Thus, an older-looking face may not just be a cosmetic issue but could reflect underlying cellular and molecular aging.
• Chronological age does not always reflect true biological aging; however, FaceAge may reveal this gap.

In summary, the inverse relationship between FaceAge and CDK6 expression supports the idea that AI-based facial age prediction may capture not only clinical but also biological and molecular aging processes.

🔍 Clinical Significance

1. Objectifies: Adds a standard metric to the physician’s subjective first impression.
2. Improves stratification: Provides independent prognostic information across multiple cancer groups.
3. Supports decision-making: Enhances prediction performance, especially in palliative care planning.

⚠️ Limitations & Ethical Concerns

📝 Conclusion

FaceAge has the potential to generate a clinically meaningful biomarker from a readily accessible facial photograph. With larger and multi-center validations, it can be safely integrated into clinical decision-support workflows.