Toward Safe Retinal Microsurgery: Development and Evaluation of an RNN-Based Active Interventional Control Framework.

Area of Science:

• Ophthalmology
• Robotics
• Surgical Technology

Background:

• Robotics-assisted retinal microsurgery improves precision but offers only passive surgeon support, lacking intelligent active guidance.
• Existing robotic systems primarily dampen surgeon hand tremor, without actively intervening to prevent surgical errors.

Purpose of the Study:

• To introduce an active interventional control framework (AICF) for enhanced safety in robot-assisted microsurgery.
• To actively intervene in surgical operations to prevent the exertion of excessive forces on the sclera.

Main Methods:

• The AICF integrates a steady-hand eye robot, a sensorized tool for force measurement, a recurrent neural network for event prediction, and a variable admittance controller.
• A recurrent neural network analyzes time-series sensor data to predict adverse events.
• A variable admittance controller actively guides the robot away from predicted hazardous situations.

Main Results:

• User studies with 14 participants, including surgeons, evaluated the AICF against auditory feedback (AF) and real-time force feedback (RF).
• The AICF demonstrated a significant reduction in undesired events, achieving approximately 2.5% occurrence.
• Benchmark approaches showed substantially higher rates of undesired instances: 38.4% for AF and 26.2% for RF.

Conclusions:

• The AICF effectively predicts and prevents excessive force application, augmenting user performance in robot-assisted microsurgery.
• This intelligent framework enhances operational safety by minimizing undesired events during critical surgical procedures.