Canonical text (the exact bytes fingerprinted)

TRUVACE RECORD VERSION
record: TRV-2026-1812
version: 1
kind: certified
reason: Certified into the record
timestamp: 2026-10-18T10:42:18.686371Z
status: published
lens: g_space
sector: health
headline: Multimodal AI refines EAU risk stratification in patients undergoing radiotherapy for prostate cancer
dek: Abstract The European Association of Urology (EAU) risk groups guide treatment decisions in prostate cancer using prostate-specific antigen (PSA), ISUP grade, and clinical stage, but do not incorporate histopathological image information. We evaluated a locked multimodal artificial intelligence (AI) model that integrates routinely available clinical variables (age, PSA, and ISUP grade) with deep-learning features extracted from diagnostic biopsy whole-slide images for predicting outcomes after radiotherapy (RT).…
gain_title: Adding a multimodal AI score to EAU classification improved model fit and separated outcomes within EAU risk groups in 353 radiotherapy-treated men.
problem_title: (none)
trace_subject: (none)
gain_reading: Adding a multimodal AI score to EAU classification improved model fit and separated outcomes within EAU risk groups in 353 radiotherapy-treated men.
gain_evidence: adding the continuous AI score to EAU classification improved model fit for BCR, HR-BCR, and PCSM
problem_reading: (none)
problem_evidence: (none)
quick_read: Researchers took a locked AI model that combines age, PSA, ISUP grade and deep-learning features from biopsy slide images. It had been trained on outcomes after prostatectomy, and they applied it without retraining to 353 men from the STHLM3 trial who received radiotherapy, then compared it with EAU risk groups.

The result suggests slide images may add prognostic information that current risk groups miss, which could eventually help individualize treatment. However, the AI score did not significantly beat EAU groups in paired comparisons, the subgroup analysis was exploratory, and prospective validation is still needed.
limitation: The AI score's improvement over EAU groups was not statistically significant in paired C-index comparisons, the risk stratification was exploratory, and prospective validation is needed.
tag: Evidence-backed gain
key_points: The model was trained on prostatectomy outcomes and applied unchanged to radiotherapy patients. | C-index for the AI score was 0.65 for BCR, 0.84 for HR-BCR and 0.81 for PCSM, versus 0.60, 0.82 and 0.80 for EAU groups. | Paired C-index differences between AI and EAU were not statistically significant. | 18.9% of EAU intermediate-risk patients were classed as AI high-risk, and 28.2% of EAU high-risk patients as AI low-risk.
rundown: The EAU risk groups use PSA, ISUP grade and clinical stage to guide prostate cancer treatment but do not use histopathological image information. The study tested whether adding biopsy whole-slide image features could refine prognosis after radiotherapy.

Endpoints were biochemical recurrence by the Phoenix definition, high-risk biochemical recurrence, and prostate cancer-specific mortality. Performance was assessed with Harrell's C-index and cumulative incidence analyses.

Within the EAU high-risk group, the 24 patients classed as AI low-risk had no biochemical recurrence over 10 years. Among the 61 AI high-risk patients, 10-year cumulative incidence was 28.7% (95% CI 17.3-41.2%).
sources:
- peer_reviewed | Scientific Reports | https://doi.org/10.1038/s41598-026-76327-w | 2026-10-17
prev: 0000000000000000000000000000000000000000000000000000000000000000

sha256

6bd918047e4b184aefbcf8fb9519bafd1c34462bc456d04d7a495c09f5dcbf07

previous

0000000000000000000000000000000000000000000000000000000000000000