Hacia la imagenología tomográfica de mama

Transcription

1 Hacia la imagenología tomográfica de mama Futuro y presente Ioannis Sechopoulos, Ph.D., DABR Advanced X ray Tomographic Imaging (AXTI) Lab Department of Radiology and Nuclear Medicine Radboud University Medical Center and LRCB Dutch Reference Centre for Screening Nijmegen, the Netherlands advanced x-ray tomographic imaging 1

2 2

3 Towards Tomographic Breast Imaging 2 D 2+ D 2.2 D 3 D Standard Mammography Stereoscopic Mammography Digital Tomosynthesis Dedicated Breast CT (If your optical system can handle it!) (Is more always better?) Digital Mammography Improvements Digital Mammography Improvements Detection in some patient subgroups (DMIST) Workflow 3

4 Digital Mammography Improvements Digital Mammography Improvements Contrast enhanced imaging Advanced imaging techniques 29% of cancers missed by overlying tissue STEREOSCOPIC DIGITAL MAMMOGRAPHY Birdwell et al, Radiology 219, (2001). 4

5 Translated X ray source X ray beam limits 6 Lesions of Interest Detector These two images are shown separately to each eye Benign Mass Half silvered mirror Cross polarized images Note shift Cross polarized lenses 5

6 High Risk Screening Study N = 1298 cases Mammo Stereo Sensitivity 12/19 13/19 Recall rate 12.9% 9.6%* Stereoscopic Mammography Promising Results D Orsi et al, Radiology, 2013 Stereoscopic Mammography Cheap! (and df fast!) Stereoscopic Mammography Too little, too late? 6

7 Stereoscopic Mammography DIGITAL BREAST TOMOSYNTHESIS Translated X ray source Recall X ray beam Lesions of Interest Detector 27 This information is used to reconstruct the volume 28 Courtesy of Hologic Inc. 7

8 Recall CC view.idc Courtesy of Hologic Inc. 30 Courtesy of Hologic Inc. FFDM System Tomo System Find the 8 differences Courtesy of Hologic Inc. 31 8

9 Benefits Mammography++ System Workflow Interpretation Dose but with some discrimination of vertical position! 33 (Some) Screening Trials Trial il N Cancer Det Recall Rate Rate STORM 2 9,672 DM: +35% Synth: +40% DM: +16% Synth: +30% OSLO 25, % 13% MALMÖ 14, % +43% slice 7/14 = 5.46 mm thick slice 36/72 9

10 An adjunct to mammography? Implementation? + DBT DM Or a replacement of mammography? Do we need two views? + CC MLO DBT DM

11 Or is one view enough? One view breast tomo: feasible? CC MLO Retrospective studies Gennaro 2010 Svane 2011 Svahn 2012 Wallis v tomo 2v DM 1v tomo 2v DM 1v tomo > 2v DM 1v tomo 2v DM Malmö Breast Tomosynthesis Screening Trial (MBTST) Lång et al v tomo > 2v DM +43 % increase in cancer detection rate Wide angle system Study design 6 readers from 3 different institutions Does 1 view experience matter? 3 readers 3 readers < 3 years > 3 years

12 Sensitivity 100% p < 0.05 * Less experienced, significant increase respect to 1v tomo, p< 0.05 Specificity 100% p < 0.05 * Less experienced, significant decrease respect to 1v tomo, p< % 90% 80% 70% 60% * * * All < 3y exp > 3y exp 80% 70% 60% * All < 3y exp > 3y exp 50% MLO DBT MLO DBT + CC DM 2v DBT + 2v DM 2v DM 50% MLO DBT MLO DBT + CC DM 2v DBT + 2v DM 2v DM Case based ROC AUC Average all readers 1v tomo not significantly different JAFROC analysis FoM NO statistical difference between 1v DBT and the rest of modalities, p = x 48 T Chakraborty et al

13 Limitations Conclusion Not screening population 1/3 were recalls seen with 2 view mammo One vendor Perhaps enough Mammogram Orig. Synthetic Tomo Slice SYNTHETIC MAMMOGRAMS Gur et al, Academic Radiology, Vol 19, No 2,

14 Recall Rates False Positive Rate DBT + FFDM % Detected Cancers DBT + Synthetic False Positive Rate % Detected Cancers 1 st Generation nd Generation Synthetic Mammograms Current synthetic 2D image can replace FFDM in combination with DBT Skaane et al, Radiology, Vol 271(3), Ratios Dose DBT/DM from Clinical Studies RADIATION DOSE 1 view DBT vs DM = view DBT vs DM = view DBT + DM vs DM = view DBT + DM vs DM = Svahn et al, The Breast, 24 (2015)

15 Synthetic (?) 1 view DBT vs DM = view DBT vs DM = TO BE ANSWERED x Svahn et al, The Breast, 24 (2015) T Remaining questions in DBT Remaining questions in DBT What is a screening DBT exam? Reading time 15

16 Remaining questions in DBT Reading strategy False negatives: mammo vs tomo Tomo + Tomo Mammo Mammo + Visibility 13 0 Radiographic appearance 3 1 Interpretative 3 6 error Lång et al, Br J Radiol 2014;87: Remaining questions in DBT Remaining questions in DBT What after a tomo screen? Multiple rounds? 16

17 DEDICATED BREAST COMPUTED TOMOGRAPHY 17

18 Albion 2003 Bodega 2007 Courtesy of Koning Corp. Cambria 2011 Courtesy of John Boone, Ph.D. Doheny 2014 Fully 3D SPECT CT: CT source down Fully 3D SPECT CT: CT source up Spiral BCT White arrow points to 16x20cm 2 CZT based SPECT camera; orange arrows point to x ray CT source, with opposed 40x30cm 2 flat panel detector; phantom on radiopaque bed. Courtesy of Dr. Martin Tornai of the Duke Multi Modality Imaging Lab Kalender et al, Eur Radiol,

19 Coronal CLINICAL IMAGES Transverse 19

20 BCT Mammo Comparison Mammo BCT: Fibroadenoma He et al, Eur Radiol, 2016 Mammo BCT: Fibroadenoma Mammo BCT: Microcalcifications He et al, Eur Radiol, 2016 He et al, Eur Radiol,

21 Mammo BCT: Microcalcifications BREAST DOSE He et al, Eur Radiol, 2016 AGD Equivalent to 2 view Mammo Breast CT vs. Mammography Boone et al, Medical Physics, 2005; 32(12), 3767 Sechopoulos et al, Medical Physics, 2010; 37(8),

22 Breast CT vs. Mammography BODY DOSE O Connell et al, J Clinic Imaging Science, 2012 Anthropomorphic Phantom Dose Variation with Projection Angle Dose Ratio (%) Lung (IL) Breast (CL) Stomach Projection Angle (deg) Sechopoulos et al, Radiology, 2008 Sechopoulos et al, Radiology,

23 Relative Organ Dose Organ 50 kvp 80 kvp Heart 1.75%/0.79% 3.08%/1.58% Lung (IL) 1.79%/2.03% 2.93%/3.25% Thymus 1.27% 2.35% Thyroid 0.08% 0.19% Uterus/Fetus 0.010% 0.026% Clavicle (IL) 1.57% 2.80% Rib Cage 4.14% 5.56% Sternum 5.16% 7.74% CLINICAL PERFORMANCE Sechopoulos et al, Radiology, 2008 First Results (N=69) Preliminary Results Diagnostic Work up 16 BI RADS 4 or 5 after work up Bx: 8 malignant lesions 8 benign lesions Breast CT Mammo 15 adequate BCT studies x Lindfors et al, Radiology, 2008 T 23

24 Preliminary Results Diagnostic Work up Breast CT: 8 malignant 8 biopsy 8 benign 5 biopsy 3 no biopsy All patients (N = 442 lesions) Sensitivity MG US BCBCT C 93/ % ( ) 93/ % ( ) 97/ % ( ) 270/ / / % 86.7 % 84.0 % Specificity ( ) ( ) ( ) AUROC ( ) ( ) ( ) He et al, Eur Radiol, 2016 BI RADS c and d density breasts (n = 270 lesions) MG US BCBCT CE BCBCT Sensitivity 78.4 ( ) 81.1 ( ) 89.2 ( ) 98.7 ( ) Specificity 70.1 ( ) AUROC ( ) 82.7 ( ) ( ) 80.1 ( ) ( ) ( ) ( ) CLINICAL IMPLEMENTATION QUESTIONS He et al, Eur Radiol,

25 Questions in BCT Questions in BCT Screening or Work up? Tissue coverage? Questions in BCT 01%of 0.1% tumors in Tail of Spence (N=839) Questions in BCT Reading time? (reading strategy) Ampil et al, Anticancer Research,

26 PET / CT for dedicated breast imaging MORFOLOGIA Courtesy of John Boone Contrast Enhanced BCT Contrast Enhanced BCT Malignant benign He et al, Eur Radiol, 2016 Courtesy of John Boone, Ph.D. 26

27 Contrast Enhanced BCT Subtraction Invasive Mammary Carcinoma Pre injection Post injection Registered Subtraction 5 Courtesy of John Boone, Ph.D. Coronal SPECT CT Volume Rendered SPECT CT Towards Tomographic Breast Imaging 2 D 2+ D 2.2 D 3 D SPECT CT Standard Mammography Stereoscopic Mammography Digital Tomosynthesis Dedicated Breast CT Large arrows point to the same location of a surgically confirmed DCIS lesion in 48yo patient. Small arrow points to posteriorly located biopsy clip. Blobs posterior to breast are chest wall signals from myocardial uptake of 99m Tc sestamibi. (If your optical system can handle it!) (Is more always better?) Courtesy of Dr. Martin Tornai of the Duke Multi Modality Imaging Lab 27

28 Towards Tomographic Breast Imaging 2 D 3 D 4 D 4D Breast CT Tumor Biology Profiling Standard Mammography Dedicated Breast CT (Is more always better?) Dedicated Breast CT FUNCTIONAL!! 4D Breast CT Neoadjuvant chemotherapy treatment planning 4D Breast CT Response prediction/monitoring 28

29 Correct Quantification 4D Noise Filtering Ramamurthy et al, PMB, Courtesy of Dr. Mathias Prokop Motion Correction PreCA PreCA registered onto CA CA Motion Correction Difference to CA Images Courtesy of Dr. Marc Kachelriess Courtesy of Dr. Marc Kachelriess 29

30 Patient BCT Classification 4D Breast Phantom Caballo et al, IEEE TMI, under review Caballo et al, to be presented at RSNA 2017 Image based phenotyping pcr prediction Courtesy of Dr. Despina Kontos Courtesy of Dr. Despina Kontos 30

31 Response Monitoring FUTURO Y PRESENTE (?) Courtesy of Dr. Despina Kontos Towards Tomographic Breast Imaging 2 D 2+ D 2.2 D 3 D Standard Mammography Stereoscopic Mammography Digital Tomosynthesis Dedicated Breast CT What if CC+MLO CC+MLO+ML 31

32 Methods Methods Diagnostic work up with mammo views Screening with DBT Combo mode Results 842 work-ups after screening DBT in previous 30 days 266 ML/LM view during work-up 28 BI-RADS 3 after work-up and excluded 133 biopsied cases 105 nonbiopsied cases 11 patients excluded due to breast size 121 patients in final cohort Breast Density Non dense 42% (51/121) Dense 58% (70/121) 0 patients excluded for breast implants 106 patients excluded followup unavailable 32

33 Abnormality Calcifications 31% (38/121) Soft Tissue 64% (78/121) Both 4% (5/121) ROC results Overall 3-View DM DBT P - value ( ) ( ) ROC results Secondary No significantdifference between: Attendings vs. fellows Microcalcifications vs. soft tissue Non dense vs. dense breasts Abnormality 3-View Digital Digital Breast Mammography (%) Tomosynthesis (%) Benign recall rate (43 84) (44 76) Malignant recall rate (50 90) (59 89) P - value

34 Reading time 3 view DM: 49.4seconds DBT: 73.7 p < Limitations Biased sampling: Screening with tomo Unilateral exams No location Towards Tomographic Breast Imaging 2 D 2+ D 2.2 D 3 D Standard Mammography Stereoscopic Mammography Digital Tomosynthesis Dedicated Breast CT Courtesy of Dr. Patrice Heid 34

35 Courtesy of Dr. Patrice Heid Courtesy of Dr. Patrice Heid Courtesy of Dr. Patrice Heid Warren et al, Medical Physics,

36 CDMAM Phantom Gold disc detection Warren et al, Medical Physics, 2012 Gold disc detection QUALITY CONTROL Warren et al, Medical Physics,

37 Screening stopped: 2016: 12 times 2015: 15 times 2014: 5 times Reasons Grid lines clearly visible Compression paddle height indication off Lines caused by mechanical shocks during readout Accidental dose increase by service engineer (70%!) Courtesy of Ruben van Engen Courtesy of Ruben van Engen Thickness indication off Thickness 63 mm W/Rh 30 kv Thickness 139 mm W/Ag 39 kv Thickness indication off Thickness 63 mm Thickness 139 mm W/Rh 30 kv W/Ag 39 kv Courtesy of Ruben van Engen Courtesy of Ruben van Engen 37

38 Lines due to mechanical shocks Example of grid lines visible Courtesy of Ruben van Engen Courtesy of Ruben van Engen Example of detector problem Step 1: Optimal Mammography Courtesy of Ruben van Engen 38

39 Advancing Human Health Courtesy of Dr. Ehsan Samei Education Clinical Service Research and Development Culture Summary Clinical need to reduce superposition Stereoscopic mammography Digital breast tomosynthesis Dedicatedbreast CT Summary Perhaps. 3 view mammography 39

40 Summary Definitely. Optimized technology at peak performance 40