[BIC-announce] FW: REPARTI-CIM Seminar: Thursday Feb. 5/09 in MC437 at 3pm - CathyLaporte

[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


-----Original Message-----
From: cim-faculty-bounces at cim.mcgill.ca
[mailto:cim-faculty-bounces at cim.mcgill.ca] On Behalf Of Marlene Gray
Sent: Monday, February 02, 2009 10:10 AM
To: cim-all at cim.mcgill.ca
Cc: Annette Schwerdtfeger; Marlene Gray; Denis Laurendeau
Subject: REPARTI-CIM Seminar: Thursday Feb. 5/09 in MC437 at 3pm -
CathyLaporte

Hello everyone,

The REPARTI-CIM Seminar for this Thursday February 5, 2009, will feature
  Cathy Laporte.

Cathy is a PhD candidate with Professor Tal Arbel, Director of the
Medical Imaging Laboratory of CIM.

http://www.cim.mcgill.ca/seminar.2009-02-02.3752885023


Title: "Learning a tissue invariant ultrasound speckle decorrelation
model"


Zames Room MC437
Thursday February 5, 2009
3:00 pm
Centre for Intelligent Machines (CIM)
3480 University
McConnell Engineering Building
McGill University

Everyone is welcome!!


Abstract:

In untracked freehand 3D ultrasound (US), image content can be used to
infer the trajectory of the transducer without a position tracking
device.  Because of the finite width of the ultrasound beam, there is a
relationship between the correlation of two US images and the
out-of-plane distance between them. The nominal relationship between
image correlation and the out-of-plane separation between images can be
established from controlled scans of a speckle phantom and used to
determine out-of-plane motion in new data sets. Unfortunately, this
nominal relationship only holds under special conditions, which hold for
speckle phantoms but are only met occasionally, or loosely, in real
tissue.  I will present a method for learning the elevational
correlation length of US signals in arbitrary tissue from a set of
example synthetic US scans using sparse Gaussian process regression.
Experiments on synthetic and real imagery of animal tissue show that
this data driven approach generalises well across transducers, yielding
results of accuracy similar to the heuristic state of the art method.
Additionally, the new approach uniquely provides a measure of
uncertainty in the estimated correlation length.

_______________________________________________
perception-list mailing list
perception-list at cim.mcgill.ca
https://lists.cim.mcgill.ca/mailman/listinfo/perception-list