Digital Twin AI system outperforms baselines for personalized cancer treatment
An RL-driven digital twin simulates treatment trajectories while a safety module blocks contraindicated drugs.
A new AI framework for clinical decision support combines treatment effect estimation, patient digital twins, and reinforcement learning to recommend personalized treatments that adapt to evolving patient conditions in real time. The system is initially trained on historical medical records and operates in a continuous learning loop. A rule-based safety module monitors vital signs and blocks contraindicated treatments, while cases with strong internal model disagreement are flagged for clinician review. The approach was validated on both a synthetic clinical simulator and a real-world ovarian cancer dataset from The Cancer Genome Atlas (TCGA).
In both settings, the AI system demonstrated superior effectiveness and stability in treatment recommendations compared to standard computational baselines. The system maintains low latency and required expert consultation for only a minority of cases in experimental validation. Designed as a clinician-supervised tool, it continuously improves through practical use, offering a promising path toward safe, adaptive personalized medicine. The paper was accepted for presentation at the IEEE Engineering in Medicine and Biology Conference (EMBC) 2026.
- Framework integrates Treatment Effect (TE) estimation, patient Digital Twin (DT) simulation, and Reinforcement Learning (RL) for sequential decision-making.
- Validated on a synthetic clinical simulator and real-world ovarian cancer dataset from The Cancer Genome Atlas (TCGA), outperforming standard baselines.
- Rule-based safety module blocks contraindicated treatments; low latency with only a minority of cases requiring expert clinician review.
Why It Matters
Real-time adaptive AI with safety guardrails could make personalized treatment decisions more effective and scalable in clinical settings.