Project Participants: This ANR project is led by St├ęphane Cottin at INRIA. I am a member of AVR / ICube on aspects related to haptic feedback. InSimo intervenes on the integration aspect. David Gaucher, ophthalmic surgeon at Strasbourg University Hospital provides us with assistance and expertise on medical aspects.

Project summary: Retinal surgery is an increasingly popular procedure for the treatment of a wide range of retinal disorders. However, like all microsurgical techniques, it requires long periods of learning before being mastered. To meet a growing demand from clinicians for very realistic training,non the one hand, and the new requirements for accreditation or renewal of surgical certifications, we propose to develop a high-fidelity training system for surgery of the retina. We seek to develop a complete simulation of the eye during all stages of a retinal surgery. This implies that the physical simulation must make it possible to make errors that can generate surgical complications. This level of realism is equivalent to what is called “Level D” for flight simulators. These simulations do not currently exist in medicine. To achieve this goal, here are the aspects we are working on in this project:

  • Visual Realism: Visual feedback is essential for surgeons, especially for ophthalmic surgery. Indeed, due to the lack of tactile feedback and the small size of the instruments, the evaluation of the surgical procedures is guided mainly by means of visual returns. The reproduction of these effects is an essential condition for quality training.
  • Biomechanical realism: the eye is a complex structure. To correctly simulate movement and deformations during all stages of surgery, all components of the eye must be modeled: the sclera, the conjunctiva, the lens, the iris and, of course, the retina. We are working on modeling methods that can simulate the behavior specific to each structure, including the epiretinal membrane that is necessary to tear during the operation.
  • Haptic Realism: Like all microsurgery techniques, the surgeon relies on very subtle indicators to avoid mistakes. Although most signals are visual, the low forces that occur during retinal surgery are also important. To reproduce these forces, we need precise and fast calculations of interactions with virtual instruments, without delay or numerical damping.
  • Complications: During training, mistakes are common. Being able to manage and apprehend these errors or complications is what really differentiates experts from non-specialists. However, very few or no simulators can learn from mistakes, since simulation often stops when an error is detected. The goal is to build a complication management system that can replicate, to some extent, the impact of errors or complications on the outcome of the surgery.
  • Assessment and Learning Curves: As with any training system, the goal is 1) to provide maximum flexibility in the training approach and 2) to objectively assess the skill level of students. To achieve the first goal, the system should be as autonomous as possible, allowing students to use the simulator with minimal supervision. An objective assessment of technical skills is another important element that needs to be measured in a thoughtful way. We seek to measure whether it is possible to clearly differentiate between novices and experts, and to determine when a certain skill level is reached.

Publications: This work has resulted in two publications [CADD15, CKAB15].