MiNaRoB team has a strong interest in the design of original and efficient control laws kwnon as Visual Servoing for medical and surgical purposes. It includes visual controllers for laser surgery, OCT-based visual servoing as well as US-based visual control.
Scientific coordinator: Brahim Tamadazte – email@example.com
Surgery makes an increasing use of laser for resection or ablation for its minimal impact on the patient. However, bringing such lasers inside the body remained a challenge until the recent technological solution based on hollow fibers, which opens the way to endoluminal laser surgery. In cooperation with AS2M's CODE and SPECIMeN groups, this action focuses on the design and control of microrobotics systems that can accurately target a laser beam on soft tissues from within the body, namely using endoscopic microvision. Within the FP7 µRALP project, it is applied to the surgery of the vocal chords (phonosurgery) which already makes an extensive use of laser surgery.
Trifocal-based Visual Servoing
This work deals with development of a weakly calibrated three-view-based visual servoing control law applied to thelaser steering process. It proposes to revisit the conventional trifocal constraints governing a three-view geometry for a more suitable use in the design of an efficient trifocal vision-based control. Thereby, an explicit control law is derived, without any matrix inversion, which allows to simply prove the global exponential stability of the control. Moreover, only ‘twenty five lines of code’ are necessary to design a fast trifocal control system.
The designed visual controller for 2 DOF mirror steering can be summarized by the block diagram shown by the figure below.
The trifocal-based visual servoing was validated experimentally in both lab testbench and on human cadaver at "Hôpital de Besançon". As can be seen in the figure below (right), the controller demonstrates very interesting behaviors (decoupling and convergence).
- Path-following for laser surgery
In this work, non-holonomic control of the unicycle model is used to implement velocity-independent visual path following for laser surgery (figure below). Generally, in the case of laser surgeries, the control consists of two tasks. The first one is to ensure the laser velocity compatibility with laser-tissue interaction (longitudinal control), in order to avoid carbonization of the tissue while achieving efficient incision or ablation. The second one (lateral control) is to ensure that the laser follows the desired geometric path drawn by the surgeon on the input device (e.g., a tablet). These two tasks eventually define the trajectory (i.e., geometric path + velocity profile along the path) of the laser. However, it is not advisable to use standard trajectory tracking because the two tasks should be intuitively modifiable by the surgeon. In fact, the velocity of the laser displacements must be the same regardless of the shape, the size or the curvature of the incision path. In addition, the norm of this velocity must be adapted, by the physician, in function of the tissue to be resected (thin, thick, fragile, etc.) or the laser source (CO2 laser, Helium-neon laser, etc.). In this paper, we focus especially on the second task: laser path following using the visual feedback independently from the velocity profile.
The developed path following method was tested and validated experimentally using an experimental set-up and on human cadaver. These experiments were performed using a high frequency (1 kHz) actuated mirror and high speed camera (10 000 fps). The developed controller has shown more than satisfactory results in terms of accuracy (average accuracy lower than 0.22 pixels (10μm) with a standard deviation of 0.55 pixels (25μm), and a relative velocity distortion, repeatability and frequency (587 Hz).