This paper presents a novel approach for semi-autonomous tissue retraction in minimally invasive surgery; taking advantage of neural networks to detect the candidate flaps for retraction from depth maps of the surgical scenario. The proposed method allows to plan and execute consistent and repeatable tool trajectories to enhance the surgeons vision during navigation in the patient’s anatomy.
Aleks Attanasio
In order to achieve semi-autonomous tissue retraction in minimally invasive surgery, a U-Net is trained to extract the tissue flap profile from the scene. To this end, the 3D reconstruction of the scene is evaluated from stereo images captured by a Da Vinci endoscope. Once the geometry of the tissues is defined, the 3D position of the flaps is used to plan and execute retraction following experienced surgeons guidelines.