Mitral Valve Regurgitation is a pathological condition where the Mitral Valve does not properly work, leading to serious complications such as heart failure or arrhythmias. Nowadays, the recommended surgical approach for treating Mitral Valve dysfunction is a minimally invasive percutaneous catheter-based procedure. Despite the achieved superior safety and efficacy, these procedures are technically demanding, requiring high level of dexterity and expertise and showing a notoriously steep learning curve. Within this framework, the ARTERY project proposes an innovative solution based on an autonomous robotic catheter system empowered with an Artificial Intelligence, Augmented Reality interface for Mitral Valve repair. To design the autonomous robotic catheter, a map between the task and the actuation space needs to be defined by means of an Inverse Kinematic model. The aim of this work is to develop and compare 3 different Inverse Kinematic models: an Analytical Constant Curvature Model, a Data-driven Model based on the Gaussian Regressor Process algorithm and a Hybrid Model, given by the combination of the Analytical and the Data-Driven model. The 3 models were tested and compared, assessing their performances based on the accuracy in following different trajectories. The average Euclidean distance error between the real trajectory and the trajectory accomplished was 3.68±1.16 mm, 2.14±1.84 mm and 1.84±1.42 mm, respectively for the Analytical, Data-Driven and Hybrid model. The Analytical model did not always behave properly, largely differing from the expected robot's behaviour sometimes, since it relied on too simplifying assumptions. However, under some simpler motions, it resulted satisfactory. At the same time, relying solely on the Data-Driven model led to wrong behaviours as well, due to errors in the collected data from which it learnt. The Hybrid model, leveraging the advantages of the 2 models, promises to be a more reliable solution, allowing accurate performance in real-time automated surgical applications and also providing a more transparent explanation about the model.
Il rigurgito della Valvola Mitrale è una condizione patologica in cui la valvola non funziona correttamente, causando gravi complicazioni come insufficienza cardiaca o aritmie. Al giorno d'oggi, l'approccio chirurgico raccomandato è una procedura percutanea minimamente invasiva basata su catetere. Nonostante l'efficacia e il crescente successo ottenuti, queste procedure sono tecnicamente impegnative e richiedono un alto livello di destrezza e competenza, mostrando una curva di apprendimento notoriamente ripida. In questo contesto, il progetto ARTERY propone una soluzione innovativa con un catetere robotizzato controllabile in modo autonomo e dotato di un'interfaccia di intelligenza artificiale e realtà aumentata per la riparazione della Valvola Mitrale. Per rendere autonomo il catetere robotizzato, è necessario definire una relazione tra lo spazio di attività e lo spazio di attuazione del robot, mediante un Modello Cinematico Inverso. L'obiettivo di questo lavoro è sviluppare e confrontare tre diversi Modelli Cinematici Inversi: un modello Analitico a Curvatura Costante, un Modello Data-Driven che sfrutta l'algoritmo di Regressione Gaussiana e un Modello Ibrido, dato dalla combinazione del modello Analitico e del modello Data-Driven. I tre modelli sono stati testati e confrontati, valutando le loro prestazioni in base all'accuratezza nel seguire diverse traiettorie. L'errore medio di distanza euclidea tra la traiettoria reale e la traiettoria realizzata è stato di 3,68±1,16 mm, 2,14±1,84 mm e 1,84±1,42 mm, rispettivamente per il modello Analitico, modello Data-Driven e modello Ibrido. Il modello analitico, basandosi su ipotesi troppo semplificative, non si è sempre comportato correttamente, discostandosi talvolta dal comportamento atteso del robot. Allo stesso tempo, anche il modello Data-Driven, in determinate condizioni, mostrava comportamenti errati, a causa di errori nei dati raccolti da cui apprendeva. Il modello Ibrido, sfruttando i vantaggi dei due modelli, promette di essere una soluzione più affidabile, permettendo prestazioni accurate in applicazioni chirurgiche automatizzate in tempo reale e fornendo anche una spiegazione più trasparente del modello.
hybrid model for a tendon-driven steerable catheter for minimally invasive mitral valve repair
Fati, Francesca
2022/2023
Abstract
Mitral Valve Regurgitation is a pathological condition where the Mitral Valve does not properly work, leading to serious complications such as heart failure or arrhythmias. Nowadays, the recommended surgical approach for treating Mitral Valve dysfunction is a minimally invasive percutaneous catheter-based procedure. Despite the achieved superior safety and efficacy, these procedures are technically demanding, requiring high level of dexterity and expertise and showing a notoriously steep learning curve. Within this framework, the ARTERY project proposes an innovative solution based on an autonomous robotic catheter system empowered with an Artificial Intelligence, Augmented Reality interface for Mitral Valve repair. To design the autonomous robotic catheter, a map between the task and the actuation space needs to be defined by means of an Inverse Kinematic model. The aim of this work is to develop and compare 3 different Inverse Kinematic models: an Analytical Constant Curvature Model, a Data-driven Model based on the Gaussian Regressor Process algorithm and a Hybrid Model, given by the combination of the Analytical and the Data-Driven model. The 3 models were tested and compared, assessing their performances based on the accuracy in following different trajectories. The average Euclidean distance error between the real trajectory and the trajectory accomplished was 3.68±1.16 mm, 2.14±1.84 mm and 1.84±1.42 mm, respectively for the Analytical, Data-Driven and Hybrid model. The Analytical model did not always behave properly, largely differing from the expected robot's behaviour sometimes, since it relied on too simplifying assumptions. However, under some simpler motions, it resulted satisfactory. At the same time, relying solely on the Data-Driven model led to wrong behaviours as well, due to errors in the collected data from which it learnt. The Hybrid model, leveraging the advantages of the 2 models, promises to be a more reliable solution, allowing accurate performance in real-time automated surgical applications and also providing a more transparent explanation about the model.File | Dimensione | Formato | |
---|---|---|---|
2023_05_Fati_Tesi_01.pdf
solo utenti autorizzati dal 11/04/2024
Descrizione: Testo Tesi
Dimensione
3.73 MB
Formato
Adobe PDF
|
3.73 MB | Adobe PDF | Visualizza/Apri |
2023_05_Fati_Executive Summary_02.pdf
solo utenti autorizzati dal 11/04/2024
Descrizione: Executive Summary
Dimensione
8.97 MB
Formato
Adobe PDF
|
8.97 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/10589/204531