An AI face-based model reported in The Lancet Digital Health can estimate biological age and predict cancer survival probability from a single photograph, using attention mapping to identify the facial regions driving the risk scores. In an internal test, 64 of 122 patients in the highest risk group died within 90 days of the photo; a second test set with more than 38,000 patients showed that higher risk-of-death scores tracked with mortality. The study frames the approach as an objective alternative to clinician “impression” judgments by learning patterns from more than 40 million face images. It also attempts interpretability by mapping where the model focuses during age and survival estimation. Separately, a Nature Communications report described a blood test for Alzheimer’s stage using plasma eMTBR-tau243 and p-tau205, claiming accuracy comparable to PET imaging for staging and predicting cognitive decline. The pairing of two plasma markers aimed to capture a later tau aggregation phase. Together, the two studies highlight a shift from purely diagnostic AI toward prognostic staging and risk-layering—moving image- and blood-based models closer to decision support use cases where clinicians need longitudinal meaning rather than one-time detection.