Deep learning models using PSMA-PET scans predict survival in metastatic prostate cancer more accurately than established clinical tools.
Key Details
- 1Researchers analyzed 486 PSMA PET scans from metastatic prostate cancer patients (2017-2025).
- 2Two AI models were developed: an imaging-only model and a fusion model combining images and clinical data.
- 3Both models were compared to six established risk tools, including VISION nomogram, CHAARTED, and LATITUDE.
- 4Fusion model achieved AUC of 0.86 and imaging model 0.82; comparators ranged from 0.43 to 0.65.
- 5Fusion model clearly stratified two-year survival (93% vs. 43%) and performed well across radiotracers and histories.
- 6Findings suggest the deep learning biomarkers could enable earlier, more personalized treatment.
Why It Matters
This study demonstrates the potential of AI-driven PET imaging biomarkers to outperform traditional risk models in stratifying metastatic prostate cancer patients, which could lead to better treatment selection and improved outcomes once validated prospectively.
Source
AuntMinnie
Related News
•AuntMinnie
AI on Chest X-Rays Matches PREVENT Score for Cardiac Risk
Deep learning models analyzing chest x-rays can predict cardiovascular risk on par with the PREVENT clinical score.
•AuntMinnie
AI Boosts Sensitivity but Raises False Alarms in Microbleed Detection
AI significantly increases sensitivity for detecting cerebral microbleeds on brain MRI but at the cost of more false positives, highlighting the complementary strengths of AI and radiologists.
•AuntMinnie
AI Algorithm Enhances Detection of Incidental Intracranial Aneurysms on Neck CT
An AI algorithm helped detect previously missed intracranial aneurysms in routine contrast-enhanced neck CT scans.