Subj:	Seminar

Invited Speaker:  Dr. Anna Jerebko, NIH/CC/DRD
Title:  Computer aided detection of colonic polyps in CT colonography"
Date:  Thursday, 8/29
Time:  12pm
Place:  Doppman Conference Room (1C520 - previously called large radiology
conference room)

Abstract:

A multi-network decision classification scheme for colonic polyp
detection and a novel segmentation algorithm are presented.

The classification approach is based on the results of voting over several
neural networks or Support Vector Machine (SVM) classifiers using variable
subsets selected from a general set. We used 21 features including region
density, Gaussian and mean curvature and sphericity, lesion size, colon
wall thickness, and their means and standard deviations. The subsets of
variables are weighted by their effectiveness calculated on the basis of
the training and test sample misclassification rates. The final decision
is based on the majority vote across the networks and takes into account
the weighted votes of all nets. This method reduces the false positive
rate by a factor of 1.7 compared to single net decisions. The overall
sensitivity and specificity rates reached are 83% and 95% respectively.
Back propagation neural nets trained with the Levenberg-Marquardt
algorithm and SVM with third degree polynomial kernel were used. Smoothed
leave-one-out validation is applied to better estimate the true error
rates.

A new polyp segmentation algorithm employing the Canny edge detector and
Radon transformation successfully segmented a selection of polyps having a
variety of shapes. Shape invariance is an important characteristic of a
successful polyp segmentation algorithm due to the wide spectrum of polyp
shapes. Our method is able to quantitatively measure particular features
of the polyp independently of its orientation and shape and use them for
classification.  This polyp segmentation method achieves high sensitivity
while reducing the number of false positive detections.