Nowadays, minimally invasive surgery (MIS), which is a set of surgical techniques that aims at reducing trauma to anatomical structures, has become a viable alternative to open surgery in many areas. In the robotic neurosurgery field, this approach is used in procedures such as biopsy, deep brain stimulation, stereoelectroencephalography and local drug delivery. This surgical approach has led to the creation of specific probes that allow to connect an entry point (EP) with a target point (TP). These probes can be divided into straight probes and steerable devices. The latter are of increasing interest because the use of the first ones may not always be optimal due to the presence of structures that prevent the feasibility of straight trajectories between EP and TP. The work presented in this thesis has been developed at the Neuroengineering and Medical Robotics laboratory (Nearlab), department of "Elettronica, Informazione e Bioingegneria" (DEIB, Politecnico di Milano), and is part of EDEN2020 (Enhanced Delivery Ecosystem for Neurosurgery ) project. Among its various goals, EDEN2020 project aims at developing steerable probes that can remain in-situ for long periods for the treatment of chronic neuro-oncological diseases. The main technology of this project, currently under development, is a PBN (Programmable Bevel-tip Needle) composed of four axially interlocked segments. These robotic segments are activated at the base of the probe so that, sliding one on the other, they can generate an offset which determines the degree of curvature at the needle's tip. The herein proposed solution aims at developing a new three dimensional path planner for neurosurgical PBN applications. Furthermore, the anatomical structures taken into account during the path planning are obtained from the patient's diagnostic images. The path planner proposed in this thesis is based on a series of algorithms run one after the other. Given ET and TP, selected by the surgeon, an entry area is generated (i.e. more EPs are highlighted in the neighbor of the first one) and subsequently the work space is limited taking into account the kinematic constraints of the PBN. Then, piecewise linear paths are generated through an RRT* algorithm, which is an improved version of the classic RRT (Rapidly-Exploring Random Trees). These paths undergo a NURBS interpolation (Non-Uniform Rational Basis Splines), integrated into a genetic algorithm. The latter aims at optimizing the smooth trajectories generated taking into account the distance from obstacles, the curvature value and the path length. Finally, the best solution in terms of a specific cost function is proposed to the surgeon. The cost function takes into account the mean and minimum distance from the obstacles, the maximum curvature and its standard deviation and the path's length. Several tests has been carried out in order to evaluate the novel proposed path planner's performances. In particular, the performances of the herein presented solution has been compared with those of two algorithms proposed in literature in the path planning field. The results show encouraging performances regarding all the parameters of interest.
La chirurgia mininvasiva, insieme di tecniche atte a ridurre al minimo il trauma arrecato alle strutture anatomiche, è diventata negli ultimi anni un'alternativa valida alla chirurgia tradizionale in molte aree. In particolare, nel campo della neurochirurgia robotica, tale approccio viene applicato ad esempio a procedure come la biopsia, la stimolazione profonda cerebrale, la stereoelettroencefalografia e la somministrazione locale di farmaci. Questo approccio chirurgico ha portato alla creazione di particolari sonde che consentano di congiungere un entry point (EP) con un target point (TP). Le sonde impiegate nella chirurgia mininvasiva possono essere divise in sonde rettilinee e sonde orientabili lungo traiettorie curvilinee. Quest'ultime sono di crescente interesse poichè l'utilizzo delle prime potrebbe essere non sempre ottimale qualora la presenza di strutture tra EP ed TP impedisca la fattibilità di traiettorie rettilinee. Il lavoro esposto in questa tesi è stato sviluppato presso il laboratorio di Neuroingegneria e Robotica Medica (Nearlab), dipartimento di "Elettronica, Informazione e Bioingegneria" (DEIB, Politecnico di Milano), e si inserisce nel progetto EDEN2020 (Enhanced Delivery Ecosystem for Neurosurgery). Tale progetto si pone, tra i suoi vari obiettivi, quello di realizzare sonde orientabili che possano rimanere in-situ per lunghi periodi per il trattamento di malattie neuro-oncologiche croniche. La tecnologia principale di tale progetto, attualmente in fase di sviluppo, è una sonda del tipo PBN (Programmable Bevel-tip Needle), composta da quattro segmenti assialmente collegati. Questi segmenti sono azionati automaticamente alla base della sonda in modo che, scorrendo uno sull'altro, possano generare un disallineamento, sulla prossimità distale, che determina il grado di curvatura nello spazio. La soluzione presentata intende realizzare un nuovo pianificatore di traiettorie tridimensionali per applicazioni neurochirugiche eseguite tramite sonda PBN. Inoltre, le strutture anatomiche su cui viene eseguita la pianificazione sono ricostruite a partire dalle immagini diagnostiche del paziente. Il pianificatore proposto in questa tesi si basa su una serie di algoritmi eseguiti in cascata. Dati ET e TP, selezionati dal chirurgo, viene generata un'area di ingresso (i.e. vengono evidenziati ulteriori EP nell'intorno del primo) e successivamente lo spazio di lavoro viene limitato tenendo conto dei limiti cinematici imposti dalla sonda. In seguito, vengono generati dei percorsi lineari a tratti tramite un algoritmo RRT*, che consiste in un'evoluzione del classico RRT (Rapidly-Exploring Random Trees). Tali percorsi vengono quindi sottoposti ad un'interpolazione NURBS (Non Uniform Rational Basis Splines), integrata in un algoritmo genetico. Quest'ultimo intende ottimizzare le traiettorie generate in termini di distanza dagli ostacoli, valore di curvatura e lunghezza del percorso. Infine, la soluzione migliore in termini di funzione di costo viene proposta al chirurgo. La funzione di costo tiene conto contemporanemente della distanza media e minima dagli ostacoli, della curvatura massima e della sua deviazione standard e della lunghezza del catetere. Per la validazione di questo progetto, sono stati eseguiti diversi test di confronto tra lo stato dell'arte ed il nuovo pianificatore. In particolare, le prestazioni della soluzione presentata sono state confrontate con quelle di due algoritmi proposti in letteratura nell'ambito della pianificazione di traiettorie. I risultati hanno evidenziato prestazioni incoraggianti riguardo tutti i parametri di interesse.
Novel methods for steerable needle 3D path planning based on NURBS and genetic algorithm
MURETTI, FEDERICO
2017/2018
Abstract
Nowadays, minimally invasive surgery (MIS), which is a set of surgical techniques that aims at reducing trauma to anatomical structures, has become a viable alternative to open surgery in many areas. In the robotic neurosurgery field, this approach is used in procedures such as biopsy, deep brain stimulation, stereoelectroencephalography and local drug delivery. This surgical approach has led to the creation of specific probes that allow to connect an entry point (EP) with a target point (TP). These probes can be divided into straight probes and steerable devices. The latter are of increasing interest because the use of the first ones may not always be optimal due to the presence of structures that prevent the feasibility of straight trajectories between EP and TP. The work presented in this thesis has been developed at the Neuroengineering and Medical Robotics laboratory (Nearlab), department of "Elettronica, Informazione e Bioingegneria" (DEIB, Politecnico di Milano), and is part of EDEN2020 (Enhanced Delivery Ecosystem for Neurosurgery ) project. Among its various goals, EDEN2020 project aims at developing steerable probes that can remain in-situ for long periods for the treatment of chronic neuro-oncological diseases. The main technology of this project, currently under development, is a PBN (Programmable Bevel-tip Needle) composed of four axially interlocked segments. These robotic segments are activated at the base of the probe so that, sliding one on the other, they can generate an offset which determines the degree of curvature at the needle's tip. The herein proposed solution aims at developing a new three dimensional path planner for neurosurgical PBN applications. Furthermore, the anatomical structures taken into account during the path planning are obtained from the patient's diagnostic images. The path planner proposed in this thesis is based on a series of algorithms run one after the other. Given ET and TP, selected by the surgeon, an entry area is generated (i.e. more EPs are highlighted in the neighbor of the first one) and subsequently the work space is limited taking into account the kinematic constraints of the PBN. Then, piecewise linear paths are generated through an RRT* algorithm, which is an improved version of the classic RRT (Rapidly-Exploring Random Trees). These paths undergo a NURBS interpolation (Non-Uniform Rational Basis Splines), integrated into a genetic algorithm. The latter aims at optimizing the smooth trajectories generated taking into account the distance from obstacles, the curvature value and the path length. Finally, the best solution in terms of a specific cost function is proposed to the surgeon. The cost function takes into account the mean and minimum distance from the obstacles, the maximum curvature and its standard deviation and the path's length. Several tests has been carried out in order to evaluate the novel proposed path planner's performances. In particular, the performances of the herein presented solution has been compared with those of two algorithms proposed in literature in the path planning field. The results show encouraging performances regarding all the parameters of interest.| File | Dimensione | Formato | |
|---|---|---|---|
|
2018_10_Muretti.pdf
solo utenti autorizzati dal 12/09/2019
Descrizione: Thesis text
Dimensione
6.96 MB
Formato
Adobe PDF
|
6.96 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/142679