Advanced Visualization Sponsored by TeraRecon

HomeCommunitiesCasesCMEAuntMinnieEuropeMiddle EastForumsConferencesMoreSign Out

TeraRecon's iNtuition enhances the clinical end-user experience provided by PACS, VNA, imageprocessing and image acquisition systems.

From sharing time-sensitive imaging data more effectively, toexpanding the clinical tools available when and wherephysicians are working; TeraRecon is an innovator and marketshare leader in advanced image management and post-processing solutions.

Ingrid Reiser, PhD,from the University ofChicago.

CAD for breast CT is improved withspiculation detectionBy Eric Barnes, AuntMinnie.com staff writer

University of Chicago researchers significantly improvedthe performance of their computer-aided detection(CAD) algorithm for breast CT studies by having itdetect lesion spiculation automatically, according to anew study.

The researchers examined spiculation detection alongwith other textural- and morphology-based features foridentifying biopsy-proven malignancies. They believetheir findings could help breast CT develop into aroutinely used modality for both screening anddiagnosis.

"Potential advantages [ofbreast CT] include the fullresolution of the 3D breaststructure, and tissue overlapis virtually eliminated," saidIngrid Reiser, PhD, in apresentation at the recentRSNA 2014 meeting."Dedicated breast CTspiculation is an importantindicator of malignancy."

Spiculation can often be seenunaided on mammography,but the task becomes morecomplicated with dedicatedbreast CT. The large volume of breast CT datasets ischallenging, and each case takes much longer tointerpret than mammography, Reiser explained.Typically, each CT dataset contains hundreds of slices,

compared with standard two-view mammography.

To address this issue, the University of Chicagoresearchers are developing CAD and diagnostictechniques to help interpret such large image volumes,said Reiser, an assistant professor of radiology. Shepresented the work on behalf of lead investigator Hsien-Chi Kuo, PhD.

The 3D spiculation feature was added to improve CAD'sclassification performance. The classification featurebuilds upon previous work by Kuo et al, which is nowincorporated into the process, Reiser said.

Use of the spiculation feature begins with manualidentification of lesion centers. Volume-of-interestextraction with image smoothing is then applied. Thetissue segmentation identifies whether the tissue isfibroglandular or adipose.

"Then we look for contact points between thefibroglandular tissue and lesion segmentation," Reisersaid. "The spiculation index is the count of connectedcontact regions between the lesion segmentation andthe tissue map. We hope that benign lesions have fewerconnected regions."

Segmentation of spiculated breast lesion image. Red(left) represents the segmented lesion (expanded forillustration), while purple (anything that appears whitein the CT scan) represents tissue classified asfibroglandular. When there's a spiculation (gray atright, or appearing as purple tissue at left), it "pierces"the red surface because it is connected to the lesion.The spiculation index counts how often the red surfaceis pierced. Images courtesy of Ingrid Reiser, PhD.

The researchers tested the feature on a database thatincluded 129 masses -- 49 benign and 80 malignant --in breast CT cases acquired at the University of

California, Davis.

Evaluating the spiculation index (SI) across thedatabase, the researchers found that SI was higher formalignant lesions (6.96 ± 3.93) than benign lesions(4.83 ± 3.34). The difference was statistically significant(p = 0.02).

Then came the feature analysis process, which included53 textural and morphologic features in all. The studyteam employed stepwise feature selection with a leave-one-out loop to assess which features were moreimportant. Receiver operator characteristics (ROC)analysis was used to distinguish malignant from benignlesions.

The area under the ROC curve for CAD improved from0.81 without the spiculation feature to 0.85 with thefeature in the task of breast mass classification (p <0.001), the researchers found.

The study had a few limitations, including the moderatesize of the database, and the fact that even though aleave-one-out evaluation method was used, thespiculation feature was developed and tested on thesame dataset. More research is needed to address therobustness of the new feature on unknown cases -- andacross different breast CT scanners and parameters.

"Our proposed 3D spiculation feature was able tosignificantly improve the performance of breast massclassification on dedicated breast CT," Reiserconcluded.

Related Reading

CAD with digital mammo doesn't improve radiologistperformance, October 8, 2014

Research sheds light on why MRI CAD misses cancers,February 3, 2014

CAD for breast MRI improves sensitivity; accuracy notso much, May 24, 2013

Breast CAD's effectiveness questioned again in newstudy, April 15, 2013

CAD ineffective for detecting more breast cancer, studyshows, July 27, 2011

About Us Advertising AuntMinnieEurope Career Center Cases Case Archives

CME Communities Conferences Contact Us Facebook Forums Home

Links Marketplace Media Kit Middle East More Reference RSNA 2014 News

Sign Out Twitter User Options Vendor Connect

Copyright © 2015 AuntMinnie.com. All Rights Reserved.

Copyright © 2015 AuntMinnie.com

Last Updated np 1/12/2015 10:40:01 AM