[BIC-announce] FW: ANNOUNCEMENT - Oral Defense for Ms Catherine Laporte

[Jennifer Chew, Ms.]

FOR YOUR INFORMATION.  JENNIFER

Jennifer Chew
McConnell Brain Imaging Centre
MNI - WB317
3801 University Street
Montreal, Qc  H3A 2B4
Telephone:  514-398-8554
Fax:  514-398-2975



________________________________
From: Grad Ece [mailto:grad.ece at mcgill.ca]
Sent: Wednesday, November 04, 2009 2:55 PM
To: ECE Graduate Students; ECE Teaching Staff
Cc: Marlene Gray, Ms.
Subject: ANNOUNCEMENT - Oral Defense for Ms Catherine Laporte

The thesis defense of Ms Catherine Laporte has been scheduled to take place Wednesday, November 11, 2009 at 2:00PM in Room 603, McConnell Engineering Building.

Statistical methods for out-of-plane ultrasound transducer motion estimation

Abstract

Freehand 3D ultrasound (US) imaging usually involves moving a conventional tracked 2D US probe over a subject and combining the images into a volume for future interpretation. Tracking devices are cumbersome; thus, there is interest in inferring the trajectory of the transducer based on the images themselves. This thesis focuses on methods for the recovery of the out-of-plane component of the transducer trajectory using the predictive relationship between the elevational decorrelation of US speckle and transducer displacement. Combinatorial optimisation and robust statistics are combined to recover non-monotonic motion and frame intersections. To account for the variability of sample correlation coefficients between image patches of fully developed speckle, a probabilistic speckle decorrelation model is developed. This model quantifies the uncertainty of displacement estimates and facilitates the use of a maximum likelihood out-of-plane trajectory estimation approach which exploits the information available from multiple noisy correlation measurements. To generalise these methods to imagery of real tissue, a data-driven method is proposed for locally estimating elevational correlation length based on statistical features collected within the image plane. The relationship between the image features and local elevational correlation length is learned by sparse Gaussian process regression using a training set of synthetic US image sequences. The synthetic imagery used for learning is created via a statistical model for the spatial distribution of US scatterers which maps realisations of a 1D generalised Poisson process to a 3D Hilbert space-filling curve. With imagery of animal tissue, the learning-based approach is shown to give distance estimates more accurate than those obtained using a speckle detection filter and comparable to the state-of-the-art heuristic method.  Remaining modelling imperfections are accounted for by an iterative algorithm which extends the proposed maximum likelihood measurement fusion approach.  In this algorithm, measurement fusion and measurement selection steps based on statistical hypothesis testing alternate to establish a trajectory estimate based on measurements which agree with each other.  This approach succeeds in avoiding distance under-estimates arising from image structures exhibiting significant but uninformative correlation over long distances.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://www2.bic.mni.mcgill.ca/pipermail/bic-announce/attachments/20091104/a090af76/attachment.htm