Enhancing Medical Training Through Learning From Mistakes by Interacting With an Ill-Trained Reinforcement Learning...

Area of Science:

• Medical Simulation
• Artificial Intelligence
• Medical Training

Background:

• Effective training for donning and doffing personal protective equipment (PPE) is crucial in healthcare, especially during pandemics.
• Traditional training methods may not adequately prepare individuals for all possible scenarios or provide objective performance assessment.
• The need for safe, remote, and efficient medical training techniques has been amplified by global health crises.

Purpose of the Study:

• To develop and evaluate a 3D medical simulation utilizing reinforcement learning (RL) and interactive reinforcement learning (IRL) for PPE donning and doffing training.
• To compare the effectiveness of RL-assisted training, which includes outlier cases, against traditional video-based training.
• To assess the impact of combining RL and IRL on participant performance and learning efficiency.

Main Methods:

• A 3D medical simulation was created with two modes: tutorial (RL agent learns PPE donning via trial and error with outlier cases) and assessment (IRL where participants provide feedback on the RL agent's actions).
• Two groups of participants were formed: one received RL-assisted training followed by IRL assessment, while the other received traditional video training before the same assessment.
• Performance was measured by the accuracy of participant feedback and the time taken for the RL agent to learn the correct PPE sequence.

Main Results:

• RL-assisted training incorporating numerous outlier cases proved more effective than traditional training focusing solely on correct sequences.
• The integration of RL and IRL significantly enhanced participant performance in the assessment.
• A remarkable 90% of participants in the RL-assisted group achieved perfect scores within three iterations, compared to only 10% in the traditional training group.

Conclusions:

• 3D medical simulations employing RL and IRL offer a superior method for teaching and assessing PPE donning and doffing procedures.
• Interactive reinforcement learning provides a robust framework for evaluating user proficiency in guiding AI agents.
• This innovative training approach demonstrates substantial improvements in learning outcomes and skill acquisition, particularly in high-stakes medical environments.