Uploaded August 2026 | Updated September 2026, 5 hours ago
PalpaRL investigates how reinforcement learning can support tumor detection by palpation in minimally invasive surgery.
Instead of probing tissue randomly, the AI learns where to palpate next by combining tactile information gathered from previous interactions. The approach was developed in a physically realistic simulation using NVIDIA Isaac Lab and realistic tissue models provided in collaboration with VirtaMed.
Compared to random and heuristic search strategies, the learned policy localizes hidden tumors with significantly fewer palpations while maintaining excellent detection performance.
This bachelor project demonstrates how reinforcement learning, surgical simulation, and medical robotics can be combined to support future minimally invasive interventions.
PalpaRL investigates how reinforcement learning can support tumor detection by palpation in minimally invasive surgery.
Instead of probing tissue randomly, the AI learns where to palpate next by combining tactile information gathered from previous interactions. The approach was developed in a physically realistic simulation using NVIDIA Isaac Lab and realistic tissue models provided in collaboration with VirtaMed.
Compared to random and heuristic search strategies, the learned policy localizes hidden tumors with significantly fewer palpations while maintaining excellent detection performance.
This bachelor project demonstrates how reinforcement learning, surgical simulation, and medical robotics can be combined to support future minimally invasive interventions.
