WIXLIB

Digital Breast TomosynthesisSWAAPM Meeting 30 Mar 2012Jerry AA. ThomasThomas, MSMS, FAAPMFAAPM, DABRDABR, CHPCHP, DABSNMDiagnostic Medical PhysicistVia Christi Hospitals Wichita, Wichita, KS

Talk Overview Breast Cancer Statistics Screen-Film Issues Tomosynthesish i–––––NeedClinical ExamplesScientific Studies on DBT vs. FFDMHow does it work?Image Display2

Acronyms in Digital Imaging FFDM – Full Field Digital Mammography– Also called Digital Mammography DBT – Digital Breast Tomosynthesis– Also called Tomosynthesis FDA “jargon”– PMA – Pre Market Approval– 510K3

BREAST CANCERSTATISTICS Estimated cases of Breast Cancer in 2010:210,000 Estimated Deaths from Breast Cancer in2010: 40,000 LifetimeLif i RiskRi k forf BreastBCancer:C1 ini 94

BREAST CANCERSTATISTICS 30% of all cancers are Breast Cancer 17% of all cancer deaths are from BreastCancer Breast Cancer is the leading cause of deathi women 40 - 44 yearsin Only 45% of women get ScreeningMammograms5

High Quality Mammography CONTRAST for Mass Identification RESOLUTION for Calcification Identification Low Patient Dose6

7

Primary Signs of Breast Cancer Mass Calcifications Mass and Calcifications8

Secondary Signs of Breast Cancer Skin ThickeningNipple InversionAdenopathyD l i DensityDevelopingD iArchitectural Distortion9

Performance of Screen/Film S/F mammography may have a miss rate of 20 –30% for breast cancer Sensitivity decreases as breast density increases10

Disadvantages of Screen-FilmScreen Film Short dynamicyrangeg– Low contrast– Under penetrationpof dense tissue Signal strength (screen thickness)compromises image quality (image blur) Film Processing– Time requiredq– 5 to 10 min– Artifacts Film ggrain noise11

Disadvantages of Screen-FilmScreen Film Can’t enhance or alter the imageg Large amount of physical storage space Must be physically transferred; only one placeat a time Information irretrievable if lost12

DIGITAL MAMMOGRAPHY13

Rembrandt Painting“Rembrandt’s Wife”14

Rembrandt’ss WifeRembrandt15

FFDM – Clinical Advantages No “filmfilm-typetype” artifacts Can see skin line without loss of contrast Faster Image acquisition– Images are available “immediately” after exposure– Increased patient throughput– Reduce patientpdiscomfort Decrease in BIRADS Category 0 Post-ProcessingPost Processing– Avoid call-backs for under exposure16

DIGITAL MAMMOGRAPHYSummary of Benefits Improved Image CONTRASTFLEXIBLE DisplayIImproveddPPatienti THROUGHPUTEasier Image Storage17

Tomosynthesis Designed to improve detection andcharacterization of breast lesions– Non-fatty breasts Multiple projections are reconstructed Allows visual review of thin breast sections–PPotentiali l to unmaskk cancers obscuredbd bybnormal tissue above or below lesion18

Why is There a Need for Tomosynthesis? In 2D FFDM:Tissue superimposition hidespathologies in 2DTissue superimposition mimicspathologies in 2DHologic – Proprietary and Confidential

Better Sensitivity ACR Phantom insert imagedwith 4 cm cadaverous breast Phantom has low contrast fibersfibers,masses, and calcifications Overlying breast tissue obscuresobjectj visibilityyReference: Andy Smith, “Overview of BreastTomosynthesis”, Hologic20

Better SensitivityDigital Mammogram 1X doseTomosynthesis 1X doseSlice at plane of phantom insertTomosynthesisTth i showshimprovedid lowl contrastt tvisibility over digital mammography21

Lower DoseDigital Mammogram 4X doseTomosynthesis 0.5X doseSlice at plane of phantom insertTomosynthesis shows improved low contrast visibilityover FFDM, even at much lower dose22

Digital Breast Tomosynthesis (DBT)— VisualizationA DBT reconstruction- 30-80 slices parallel to thedetector plane- 1mm slice thickness- 100 µm in-plane pixel sizeVisualization software functions- Paging through DBT slices- Window level- Zoom in / zoom out- Field of view magnifierThe knowledge for interpretingconventionaltil mammographyh iisvalid for DBTCourtesy of Tao Wu, Ph. D and Massachusetts General Hospital23

Breast TomosynthesisA three-dimensionalthree dimensional mammographic examinationthat can minimize the effects of structureoverlap within the breast24

Breast Tomosynthesis Preserves the very highresolution of 2D FFDM Multiple images of the breast areacquired at different anglesduring a sweep of the x-rayx ray tube Allows radiologistsgto see aroundoverlapping structures

Hologic Selenia Dimensions DBT 2D or 3D imaging– 2D onlyl– 3D only Combo mode– 3D image– Return to 0 degreesgfor 2D imageg– Single compression for both images26

How Does Hologic’s Tomosynthesis Work? Tube moves in a 15 arc 15 low dose images are acquired 1 image at each degree Four second sweep Total dose one 2D mammogram Images are reconstructed into 1 mm slicescombo--modecombo Inimaging, 2D and3D iimages are ttakenk underd theth samecompression, with no additional patientpositioning required. Combo supports bothCC and MLO projections.projectionsCompression PaddleCompressedpBreastDetector Housing

Potential Benefits of 3D Imaging Better imaging– Improved lesion margin visibility– Precise lesion localization– Identification and location of multimulti-focalfocal cancers Higher accuracy– Increased breast cancer detection– Higher PPV for breast biopsy recommendations– Decreased workup rate for non-cancer cases LLower recallll ratest– Decreased workup rate for non-cancer cases28

Hologic ROC Study for FDAPMA29

ROC Study Design 1083 women were recruited from 5 clinicalcenters– 856 presented for screening mammography– 227 presented for breast biopsy All subjectsjreceived 2D and 3D imagesg ofboth breasts in CC and MLO positions Radiation dose for a single 2D plus 3Dacquisition (either CC or MLO) was lessthan the MQSA limit for a single 2Dmammogram30

Overview of Reader Study Comparison of 2D to 2D plus 3D Two retrospectivei Independentdd Readerd StudiesS di Readers were MQSA qualified– Wide range of experience in 2D Reader study enriched with:– CCancer cases– Recalled screening cases– Benign biopsy cases Major conclusions– Improved area under ROC curve– Reduced recall rate31

Rationale for using 2D plus 3D Comparison of current images with prior images is standardmammography practice and critical to perceive subtle changeswhich may be associated with a cancer. Obtaining a 2D exam along with the 3D exam will allow directcomparison of current 2D images with prior 2D images Segmental and clustered calcifications are more easily andquickly appreciated with 2D because they can traverse multipleslices in 3D. By minimizing structure overlap, 3D optimally demonstratesmasses and architectural distortion32

Reader Study #1 Pooled ROC Curves for12 Readers Significant increase inperformance60% increased to 78%Hologic – Proprietary and Confidential33

Reader Study 2 Pooled ROC Curves for15 Readers Identical Results from 2independent reader studies Significant increase inperformance60% increasedid tot 76%Hologic – Proprietary and Confidential34

Reader Study 1 & Reader Study 2Pooled ROC Curves Pooled ROC Curves for 2Reader Studies Almost complete overlapbetween the two studiesHologic – Proprietary and Confidential35

Pooled ROC ResultsWith three points to illustrate trade off between cancerdetection and recall rate changesCancer DetectionIncrease16%8%0%Recall RateReduction0%‐25%25%‐40%2D pplus 3D2DAUC diff 0.072, p‐value 0.0001Hologic – Proprietary and Confidential

Reader Study 2: Pooled ROC CurvesReader Study 2 added a3D MLO viewIncreased sensitivity anddecreased recall rateHologic – Proprietary and Confidential37

Performance in Dense BreastsTomo improved ROC performance in fatty breastsIn dense breasts, ROC performance increased 3X that of fattyConclusion: Tomo useful in fatty breasts, more useful in dense breastsFatty – BIRADSdensity 1 or 2Hologic – Proprietary and ConfidentialDense – BIRADSdensity 3 or 438

What are the benefits of Combo-mode* The ROC analysis demonstrated that 2D plus 3D issuperior to 2D alone The ROC results showed that for a given sensitivity,the recall rate should be lower using tomosynthesis The ROC results showed that at a given recall rate,sensitivityses v y shoulds ou d be higherg e ususingg tomosynthesiso osy es s The ROC analysis demonstrated that theperformance of all ppparticipatingpg radiologistsgimproved, regardless of experienceNone of these statements could be said for the transitionfrom Analog to Digital Mammography *The Hologic Selenia Dimensions clinical studies presented to the FDA as part of Hologic’s PMAsubmission that compared Hologic’s Selenia Dimensions combo‐mode to Hologic 2D FFDM39PRE‐00023

TTomosynthesish i TechnologyT h lPutting it all together

The technology behindtomosynthesis Underlying technologies––––Digital detectorsX-ray unitReconstruction algorithmsImage Display41

Engineering constraints Total radiation doseImaging timePatient motionDDetectorperformancefDetector motionAbility to image entire breastNeed to provide for biopsy of lesions onlydetected by DBT42

Design Approach Arc of movement– (11 – 60 Degrees) Number of projections– (9 – 25) Exposure– Continuous or pulsed Detector Exposure parametersTotal DoseEffective size of pixelsX-rayy source / filterSingle or binned pixelsPatient position– Fixed or moved43

IMAGE GENERATION44

Schematic of TomosynthesisPark J M et al. Radiographics 2007;27:S231-S24045

Projection of Objects in BreastPark J M et al. Radiographics 2007;27:S231-S24046

Reconstruction Algorithms Shift-and-addTuned Aperture CTMatrix InversionFil d backFilteredb k projectionj i (FBP)Maximum likelihood reconstruction (ML)Simultaneous algebraic reconstruction (SART)Gaussian frequency blending (GFB)Voting strategy47

Basic Principle ofSlice ReconProjection 1Shift each projection left orright then add to get the planeProjection 2Projection 348

Basic Principle ofSlice ReconShift RightShift each projection left orright then add to get the planeNo ShiftShif LeftShiftfAdd shifted images toget final slice image49

Basic Principle ofSlice ReconShift LeftShift each projection left orright then add to get the planeNo ShiftShif RightShifti hAdd shifted images toget final slice image50

Shift and AddPark J M et al. Radiographics 2007;27:S231-S24051

Image Display Single Slice Slab Recon––––ArithmeticGeometricCubicOth ?Other? MIP52

MIP of slices showing Calcifications53

MIP of slices showing speculated lesionsStatusEND54

Clinical Image TOC Better Visualization Recall Reduction– Tissue superimposition – mimicking Cancer InvasiveInasi e DuctalD ctal Carcinoma (IDC)Micropapillary type Ductal CarcinomaMetastasis from endometriod carcinomaArtifacts55

Clinical Data AcknowledgementImages and data courtesy of: Hôpital Privé d’Antony, Paris France Massachusetts General HospitalHospital, Boston MA USA Netherlands Cancer Institute –Antoni Van Leeuwenhoek Hospital, Amsterdam Holland Centre de Radiologie et dd’EchographieEchographie du Docteur Joussier,Joussier Paris France Dartmouth Hitchcock Medical Center, Lebanon NH USA Magee Women’s Hospital, Pittsburgh PA USASlides courtesyy of: Andy Smith, Ph.D. Hologic, Inc56

Better Visualization Example 1

LCCA 2D Mammography Image2DHologic – Proprietary and Confidential58

LCCA suspiciousi iarea ini a 2D MammographyMh ImageI2DHologic – Proprietary and Confidential59

The 2D Mammography Image next to one slice of a 3D Image Set2D3DThe Difference is ClearHologic – Proprietary and Confidential60Status

Recall ReductionSuperimposed Tissue Examples

A 2D Mammography Image witha suspiciousarea next to a 3DRCCimage set2DHologic – Proprietary and ConfidentialRCC3D3DClick image torun 3D in cine62

Slice3026221418Steppingpp g thru the imagegset, shows that thesuspicious area is nothingmore than normal breaststructures overlappingHologic – Proprietary and Confidential63Status

Recall Reduction – Superimposed Tissue (Case 2)Hologic – Proprietary and ConfidentialClick image torun 3D in cine64Status