University of Twente Student Theses

Login

Respiratory motion estimation of the liver with abdominal motion as a surrogate : a supervised learning approach

Fahmi, Shamel (2017) Respiratory motion estimation of the liver with abdominal motion as a surrogate : a supervised learning approach.

[img]
Preview
PDF
13MB
Abstract:Currently, imaging techniques are advancing widely in the medical field especially during image-guided interventions and diagnosis. Among the available medical imaging techniques, magnetic resonance imaging (MRI) offers the highest soft tissue contrast, which helps detecting small lesions (such as tumors) at an early stage. However, MRI does not offer high update rate acquisitions and thus the exact motion of the lesions remains uncertain. Moreover, respiratory induced motion introduces significant challenges during medical image acquisition and image-guided interventions. Respiratory induced internal liver motion causes uncertainties in localizing hepatic lesions which could lead to motion artifacts and misdiagnoses during image acquisition or inaccurate targeting and significant tissue loss in case of image guided interventions. A common approach is Respiratory Motion Estimation (RME) in which the internal liver motion is estimated by measuring external signals called surrogates that do not directly measure the internal liver motion. The aim of this thesis is to determine the feasibility of estimating the internalmotion of the liver due to respiration acquired usingMRI by tracking small sized markers using a digital camera. The two acquired data will be processed offline and a fitting algorithm will be developed to design motion models such that based solely on tracking the external markers at a high update rate, the liver motion is estimated. In the same context, three healthy subjects volunteered for human subject experiments. Each volunteer was subjected to two sessions such that MRI acquired liver images were recorded alongside with camera tracked external markers. The acquired liver and abdomen motion were utilized to train three motion models (multiv-ariate, Ridge and Lasso regression models) to estimate the superior-inferior (SI) motion of the liver. The conducted human subject experiments demonstrated that the breathing patterns differ between sessions and subjects and thus, patient specific motion models were designed. The liver SI motion estimated by the motion models were compared to the true values acquired from MRI. Over the six acquired sessions, the mean absolute error (MAE) predicted by the motion models ranged between 0.8 mm and 1.9 mm. During the period of this thesis, a medical proposal was submitted to the local medical ethical committee at the University of Twente to approve conducting the human subject experiments. The medical proposal consisted of two main documents. Firstly, a detailed measurement protocol was documented to explain the step by step procedures of the conduced human subjects experiments (Appendix [A]). Secondly, a detailed description of the study was given to the volunteers prior signing their consent (Appendix [B]). Furthermore, before conducting the human subject experiments, various preliminary experiments (Appendix C) were conducted to formulate and develop the measurement protocol. Robotics
Item Type:Essay (Master)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:44 medicine, 50 technical science in general, 54 computer science
Programme:Systems and Control MSc (60359)
Link to this item:http://purl.utwente.nl/essays/73643
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page