Abstract #1069
Prostate cancer detection from contrast enhanced T1 time course without pharmacokinetic modeling
Nandinee Fariah Haq 1 , Piotr Kozlowski 2,3 , Edward C. Jones 4 , Silvia D Chang 3 , Larry Goldenberg 2 , and Mehdi Moradi 1
1
Electrical and Computer Engineering,
University of British Columbia, Vancouver, BC, Canada,
2
Urologic
Sciences, University of British Columbia, Vancouver, BC,
Canada,
3
Radiology, University of British
Columbia, Vancouver, BC, Canada,
4
Pathology
and Laboratory Medicine, University of British Columbia,
University of British Columbia, Vancouver, BC, Canada
In this work, we propose a data-driven approach to
characterizing T1 time course. This method which is free
of physiologic modeling is used to classify prostate
tissue into cancer and normal, based on dynamic contrast
enhanced T1-weighted images. The reference standard is
the wholemount histopathologic analysis of extracted
prostate specimens. Our approach is to design a learning
agent that can detect cancer directly from the T1 time
course without modeling the physical phenomenon. The
dimensionality of the T1 time course is reduced using
Principal Component Analysis (PCA) and the resulting
parameters are used with Support Vector Machine
Classification (SVM). An area under ROC of 0.87 is
reported in pixel level classification.
This abstract and the presentation materials are available to members only;
a login is required.
Join Here
