Description du projet
Une nouvelle manière de piloter les robots chirurgicaux
Les robots chirurgicaux seront-ils un jour capables de fonctionner de manière autonome? Les machines contrôlées par ordinateur utilisées en chirurgie sont directement pilotées par des chirurgiens chevronnés. À l’heure actuelle, il n’existe aucune forme d’autonomie robotique. Bien que les approches robotiques modernes visent la précision géométrique absolue, c’est la précision fonctionnelle (relative au ciblage de structures anatomiques et fonctionnelles) qui constitue le paramètre le plus important en chirurgie. Le projet FAROS, financé par l’UE, développe actuellement de nouvelles manières d’incorporer l’intelligence physique en robotique chirurgicale. Plus spécifiquement, il compte construire une représentation fonctionnelle de la tâche chirurgicale en combinant des capacités de détection non visuelles clés. En outre, l’apprentissage automatique profond interprètera les données intraopératives. Le projet fera la preuve de cette nouvelle approche de fonctionnement des robots chirurgicaux dans le cadre d’étapes critiques exécutées de manière autonome pour une chirurgie de la colonne vertébrale.
Objectif
FAROS aims at improving functional accuracy through embedding physical intelligence in surgical robotics. A key motivation for introducing robots in operating rooms has been their ability to deliver superhuman performance. However, for the vast majority of surgical procedures, robotic positioning precision alone is not sufficient to realize the “right gesture”. Indeed, surgical accuracy is a different concept from standard engineering notions such as geometric precision, resolution or sensitivity. This arises from the essence of the surgical tasks: surgeons do not let their gestures be dictated by pure geometric objectives; rather, functional objectives are what they pursue. FAROS explores venues to efficiently embody surgeon-like autonomous behaviour at different levels of granularity. The following key ingredients are foreseen: (1) a rich set of non-visual sensors that form a multifaceted representation of the surgical task; (2) functional models that relate non-conventional sensor signals to functional parameters (e.g. tissue type, quality of tissue or bone, condition of tissue/fluid, tissue damage, perfusion, implant stability, etc.); and (3) functional controllers, obtained through reinforcement learning, that encode physical intelligence and produce sensible autonomous robot actions geared at closing knowledge gaps or optimizing functional performance. This new concept, which we refer to as Functionally Accurate RObotic Surgery (FAROS), will be showcased on two critical spine surgery use cases, namely: pedicle screw placement and endoscopic lumbar discectomy. A compact yet multi-disciplinary team consisting of academics, industry and end-users will collaborate closely to build up robotic controllers that are better suited at delivering functional accuracy in the presence of large variability and disturbances inherent to every surgical act.
Champ scientifique
- natural sciencescomputer and information sciencesartificial intelligencemachine learningreinforcement learning
- engineering and technologyelectrical engineering, electronic engineering, information engineeringelectronic engineeringsensors
- medical and health sciencesclinical medicinesurgeryrobotic surgery
Programme(s)
Régime de financement
RIA - Research and Innovation actionCoordinateur
3000 Leuven
Belgique