Performance of a Previously Validated CT Colonography Computer-Aided Detection System in a New Patient Population
Ronald M. Summers1,
Laurie R. Handwerker1,
Perry J. Pickhardt2,
Robert L. Van Uitert1,
Keshav K. Deshpande1,
Srinath Yeshwant1,
Jianhua Yao1 and
Marek Franaszek1
1 Diagnostic Radiology Department, National Institutes of Health Clinical
Center, Bldg. 10, Rm. 1C368X MSC 1182, Bethesda, MD 20892-1182.
2 Department of Radiology, University of Wisconsin, Madison, WI.
View larger version (8K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 1A —Free-response receiver operating characteristic (FROC) curves
for external validation (diamonds) and previously reported internal
validation training (triangles) and testing data sets
(squares). Curves show the trade-off in sensitivity per adenomatous
polyp versus number of false-positive findings per patient as tunable
threshold is varied in computer-aided detection (CAD) software. Circles
indicate operating points used to determine sensitivities in
Table 1. Graph shows FROC
curves for adenomatous polyps 10 mm in diameter and larger. Sensitivity of CAD
in external validation data set is not significantly different from previously
reported internal validation data sets.
|
|
View larger version (9K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 1B —Free-response receiver operating characteristic (FROC) curves
for external validation (diamonds) and previously reported internal
validation training (triangles) and testing data sets
(squares). Curves show the trade-off in sensitivity per adenomatous
polyp versus number of false-positive findings per patient as tunable
threshold is varied in computer-aided detection (CAD) software. Circles
indicate operating points used to determine sensitivities in
Table 1. Graph shows FROC
curves for adenomatous polyps 6–9 mm in diameter. Sensitivity of CAD in
external validation data set is greater than for both previously reported
internal validation data sets.
|
|
View larger version (95K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 2A —56-year-old woman with 8-mm sessile tubulovillous adenoma in
ascending colon detected with computer-aided detection. Three-dimensional
prone endoluminal virtual colonoscopic images without (A) and with
(B) blue computer-aided detection mark. Arrows in A indicate
polyp.
|
|
View larger version (96K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 2B —56-year-old woman with 8-mm sessile tubulovillous adenoma in
ascending colon detected with computer-aided detection. Three-dimensional
prone endoluminal virtual colonoscopic images without (A) and with
(B) blue computer-aided detection mark. Arrows in A indicate
polyp.
|
|
View larger version (93K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 3A —83-year-old man with 6-mm sessile tubular adenoma on side of
haustral fold in transverse colon detected with computer-aided detection.
Three-dimensional prone endoluminal virtual colonoscopic images without
(A) and with (B) blue computer-aided detection mark. Arrows in
A indicate polyp.
|
|
View larger version (94K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 3B —83-year-old man with 6-mm sessile tubular adenoma on side of
haustral fold in transverse colon detected with computer-aided detection.
Three-dimensional prone endoluminal virtual colonoscopic images without
(A) and with (B) blue computer-aided detection mark. Arrows in
A indicate polyp.
|
|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
Copyright © 2008 by the American Roentgen Ray Society.
