Automation of Surgical Tasks Using Hybrid Dynamical Systems - #VTMS

Relatore:  Michele Ginesi - Università di Verona
  martedì 9 marzo 2021 alle ore 21.00
Ciclo #VeronesiTuttiMathSeminars #SafetyThenEnjoyMath clicca per aggiungere al tuo calendario Google

Automation of Robotic Minimally Invasive Surgery holds significant promise to improving patient treatment, reduce recovery time, and reduce the fatigue of the surgeons. To achieve automation in surgical robotics, multiple challenges must be addressed (e.g. scene recognition, planning, control, etc.).
Surgeons master their skills with years of training, and gestures are performed using muscular memory and expertise.
In this seminar, we will talk about gesture learning and segmentation. We will present a Hybrid Dynamical System (HDS), i.e., a system composed of both a discrete and a continuous component. The discrete component determines the currently active mode, and each mode describes the continuous evolution of the robot's end-effector position.
For instance, the suturing task can be subdivided into many gestures (looping, knot-tying, etc.). Each gesture is a mode, and it can be achieved by performing a particular movement with the robot end-effector.
Modeling the surgical task as an HDS allows learning both the continuous evolution and the mode transition rules from observations of real surgical tasks. The continuous component of the HDS is modeled using the Dynamic Movement Primitives (DMPs) framework, which allows learning trajectories of any shape. To learn the mode transition rules, a segmentation of the surgical task into segments is needed. Manual segmentation is a tedious process, and errors are practically unavoidable. Thus, we propose an unsupervised method based upon the Auto-Regressive Hidden Markov Model.
The seminar will first introduce the problem of automation in Minimally Invasive Surgery. Then, we will present our proposed HDS model. Finally, some preliminary results in the automation of surgical-related tasks will be presented.

Il link zoom per accedere al seminario è e la password è costituita dalle prime 8 cifre significative di arccos(-1).
Per rimanere informati sui successivi seminari, si prega di iscriversi alla seguente mailing list

Marco Caliari, Presidente del Collegio Didattico di Matematica
Matteo Frigo, Presidente dell'Associazione Alumni Matematica Verona

Marco Caliari

Data pubblicazione
20 gennaio 2021

Offerta formativa