(ISMRM 2010) Improving Robustness of Cartilage Segmentation Using IDEAL Water and Fat Images
Meeting Banner
Abstract #5118

Improving Robustness of Cartilage Segmentation Using IDEAL Water and Fat Images

Raghu Kokku1

1MR SW & Apps , GTO-I, Wipro GE Healthcare, Bangalore, Karnataka, India


Accurate and reliable quantification of cartilage volume in MRI is required for diagnosis of many degenerative and inflammatory diseases such as osteoarthritis or rheumatoid arthritis. A Novel approach to segment the anatomical structures and cartilage using IDEAL knee MRI data is proposed. Variation in the characteristics of similar structures in IDEAL water and fat images is used to generate the guidance map for automated segmentation. Segmented structures are analyzed qualitatively and quantitatively with manually segmented datasets from GE 1.5T scanner. Reported DSC with the experimental datasets (>85%) indicates that the proposed solution improved the robustness of segmentation.

Keywords

cartilagesegmentationidealtissuewateranatomicalguidancestructurestibiaaroundfemurmanualproposedadiposemaskmuscularsegmentedshapeaccuracyanatomybonecharacteristicsintensitypatellavariationaccuratebackgroundboundaryclassificationdatasetsdominantfeaturesgeneratedimprovedkneemanyreederreliablerobustrobustnesssegmentsmartlysolutionsuppressionvolumeaccountacquisitionactiveanalyzedarthritisattachedautomatedaveragingbandwidthbetterboundariesbrokenchallengechallengedchemicalcoefficientconcludedcontourscreateddefinedefineddegreesdescriptiondetectiondetectsdetermineddevelopeddiagnosisdicedifficultdistributioneasilyefficiencyefficienterrorsevaluationexpectextendedfalsefieldfinalfoldgenerategeneratinggrowingimproveimprovinginconsistenciesindicateindicatinginflammatoryinfluencinginsightinspectionjennyknowledgemakesmeasuredminormodelmorphologicalmuscleneednegativenovelnullifiedoftenoperationoperatorosteoarthritisoutlinespositivesprescriptionproposeprovenquadrantsquantificationrelativelyremainingremainsreportedrequiredresolutionreviewedrheumatoidscannersearchsharesimilaritysmartspacespatialspeedstrongstrongersubjectssurroundingtakethoughtissuesusageview