Please use this identifier to cite or link to this item:
Title: Real-time monitoring and registration of brain shift using hybrid control strategy for robotic neurosurgery
Authors: Vineet Jacob Kuruvilla
Keywords: DRNTU::Engineering::Mechanical engineering::Surgical assistive technology
DRNTU::Engineering::Mechanical engineering::Bio-mechatronics
DRNTU::Engineering::Computer science and engineering::Computing methodologies::Image processing and computer vision
Issue Date: 2013
Source: Vineet Jacob Kuruvilla. (2013). Real-time monitoring and registration of brain shift using hybrid control strategy for robotic neurosurgery. Doctoral thesis, Nanyang Technological University, Singapore.
Abstract: Brain, being a soft tissue, deforms dynamically during neurosurgery. Maximum displacement of 50mm at the end of tumour resection surgery (6-7 hours) has been observed in some studies. This phenomenon of intra-operative brain tissue deformation is called brain shift. Currently available non-invasive neurosurgery robotic systems such as Cyberknife®, assumes a fixed spatial relationship between brain abnormalities like brain tumour and the skull. This can lead to inaccurate targeting due to brain shift when using non-invasive surgical modalities such as High Intensity Focused Ultrasound (HIFU). Frequent tracking of the brain shift followed by dynamic update of the treatment plan needs to be performed for positional compensation. In this research, a control framework for non-invasive neurosurgical robots, using dynamic image registration for tracking the intra-operative deformation of the brain in real-time is devised. The estimation of intra-operative deformations of the brain tissues are addressed for improving the accuracy. A two-stage dynamic deformation tracking strategy was developed in this work for the intra-operative brain shift estimation using a fast and computationally light template matching algorithm for the rigid shift estimation (gross position information) and a more computationally intensive point-based non-rigid image registration algorithm called coherent point drift algorithm (CPD) for the non-rigid shift estimation (fine position information). For the experimental testing, ultrasound imaging was used as the intra-operative imaging modality and phantoms made of agar agar gel with embedded target were used. An average error of 0.4mm and average computation time of 0.8second was obtained from template matching algorithm and average error 1.94mm with a computation time of approximately 72.7seconds for the coherent point drift algorithm. The computation time was reduced by a factor of 5 to 9 by introducing a down-sampling factor to the number of points in the registered point sets. A hierarchical hybrid supervisory control combining a supervisor, direct model reference adaptive control (DMRAC) and proportional-integral-derivative (PID) control for non-invasive neurosurgical robots is proposed. The control architecture consists of three layers- the high level supervisor, mid-level and the low level control. The PID controller and DMRAC were formulated, simulated, implemented and tested and the performances were compared. DMRAC showed 22% better accuracy compared to PID controller in workspace regions where the dynamic effects (friction, moment of inertia variations) were significant especially in the Distant Point protocol. In workspace regions or Adjacent Point protocol where the dynamic effects were minimal, both PID controller and DMRAC showed comparable performance. The control derivation of the hybrid supervisory control, its simulation and experimental testing was performed. The results showed better accuracy compared to using a standalone PID controller by switching to DMRAC when the performance of PID controller deteriorated. The accuracy of the overall system was 22% better compared to the performance of PID controller in Distant Point protocol and no significant advantage in Adjacent Point protocol. The proposed control is tested on the end-effector of a representative neurosurgical robot, FUSBOT-NS (Focused Ultrasound Surgery Robot-Neurosurgery).
DOI: 10.32657/10356/53098
Fulltext Permission: open
Fulltext Availability: With Fulltext
Appears in Collections:MAE Theses

Files in This Item:
File Description SizeFormat 
TmG0702118L.pdf20.34 MBAdobe PDFThumbnail

Page view(s) 50

Updated on Feb 25, 2021

Download(s) 50

Updated on Feb 25, 2021

Google ScholarTM




Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.