Volumetric Breast Density Analysis in Mammography and Tomosynthesis - Brief Overview Video

OK, so thank you for the kind introduction and thank you to the organizers for inviting me today. So these are the learning objectives. To understand the clinical basics of volumetric breast density analysis, VDS based on some examples, previous studies. To acknowledge the difference between radiologists, assessment breast density and software measurements. And finally, to discuss VDS potential use for mammography and digital Bristol synthesis. DBT in clinical practice. So as you know, this is mammographic density. An increased mammographic density on your right. Which is associated per, say, with an increased risk of breast cancer. And mammographic density, it's something that makes it more difficult to see what we want to see. So the white breast tumor in the white dense breast or the floats and the fishing Nets so dense. Today it is a message from radiologists to radiologist how difficult the breast is to interpret, but it's also, I think, a message to the woman. A message that should really be handled with care, since it implies something about the risk. And density is also great tool as illustrated by this nice drawing from the cover story and Council world. A great tool to be used in population screening and in personalized medicine and we will get back to that in the end. So just shortly there are different measurements of breast density. It's the qualitative ones that lets us radiologists looking at the image, and it's based on visual perception and then this, the quantitative ones and we will focus on the volumetric ones and I cannot or will not go into specific functions of the different softwares, but in general they are based on different aspects of statistical, mathematical or physical. Modeling that estimates density. So the first learning objective. Of course, the focus should be on the most important endpoint clinically, that is breast cancer risk and this study in 2014 concludes that despite differences in their density distribution, these measurements all produce positive associations with risk. Strongest football powering Cumulus, however, just this February. This was published in radiology showing that by rats the qualitative one. Classification was as accurate as computer assistance methods for discrimination of patient from control subjects. So both of the methods still adds value I think. But focusing on the volumetric methods are they measuring what we want them to measure or they valid? And are they doing so in a consistent way? So are they reliable and it should be pointed out that different papers often just of course different statistical methods of comparing such as agreement or correlation and so forth. And it sometimes makes the bit tricky to compare. Starting with validity, we tend to want to use or to compare with MRI. In some sense being closest to the truth regarding in the tissue or the the breast tissue. So has been done by these two studies from 2013 and 2014. And the top one states that volumetric measurements obtained from full field digital mammography with valpara show high correlation with MRI data. And the bottom study showed that classification of women by volumetric density by any of the three mammographic techniques is compareable to classification by MRI density. Moving to reliability, there are of course many factors affecting reliabilities, such as different positioning of the breasts between exams, different levels of compression or technical aspects, or even different nurses performing the studies and these two papers, among others. Of course I've been looking at reliability and the up once they said volumetric measures obtained using full power control algorithms have excellent reliability for repeated measures. Weather cumulus meshes have moderate reliability and these women underwent. Repeated imaging at the same same time, but we do with two technologist. The lower one is a Dutch trial and they show that the categorization of serial mammograms in density classes is more consistent with automated software. The useful power then with a mixed group of human readers and this paper is from a screening setting. So to conclude, a they seem both volumetric measurements seems valid an reliable, although we always need larger studies to validate this. Of course the focused should be on which method the best shows the Association between density and breast cancer risk, but as a radiologist I think it's still interesting to compare where the software, since if we're about to shift to a volumetric method of software, we want to know what our results translates into. And we looked at this using the malama Bristol synthesis trial that Christine along will tell you more about this afternoon and we compared radiologist scores with the software in mammography and we found the moderate agreement between volumetric density and birads. This was the 4th edition scores from radiologist and just to compare with some other large trials. This just some basic facts about the quantitative estimation of the qualitative one an the population. And an the results from those papers were to conclude moderate to substantial agreement between software and radiologists. the Apple study looked at correlation, but if the correlation is poor, the agreement is also bound to be at lower end, and we will go back to get back to potential reasons for this varying agreement further on. So I would just want to say something about tomosynthesis as well. So as you know, tomosynthesis has the ability to compliment or perhaps even replace mammography. So we want feasable tool or perhaps a software to measure density. Also in tomosynthesis and these papers looked at differences in density estimation between modalities. So between MRI, mammography and DBT. And in general they found a difference between one to 10% in estimated density between the modalities and offer. The mammography was somewhat lower. The density in the mammography was estimated somewhat lower than in the other modalities. And just As for mammography, we have been looking at differences in breast density between software and radiologist. And our research group looked like this. This is a conference proceeding from the eye WDM in 2016, and we looked at a small part from the Mama breast tomosynthesis trial, and this was research collaborations with Siemens, so we used their volumetric breast density software, which is also taking masking into account, and it's described more in detail in this publication, which was also a conference proceeding. And we saw that volumetric density in mammography versus Thomas synthesis using the central projection show high correlation and a substantial agreement. And we also wanted to compare the radiologist with the software both in mammography, Anna Thomas Synthesis, and we saw a substantial agreement between radiologists and now the Pirates 5th edition and software. The agreement was somewhat lower in the tomosynthesis image, ingane. One potential reason for that could be that we used the software use the central projection and a mammography. We have two images progressed. So such you conclude, it seems feasible to use the volumetric measurements also in terms synthesis. However, larger studies are needed in our research group are planning an extension of this project to include the whole volume of the whole population of the Mama breast imaging thesis screening trial, and that is 15,000 women. So we're looking forward to that. So it's very second learning objective to acknowledge potential differences between radiology radiologists and the software. So radiologists are looking a lot on masking so that being the message or the image is it difficult to interpret and most softwares are looking at pure volume, so that could be one potential reason for the varying agreement, as radiologists are looking at process data. So when the images taken this first for data and then there's various steps of course of image handling. So they are a technical differences between the raw data that the software usually uses and the process data that we as radiologists using. And finally, it has been shown that American radiologists tend to rate density higher than European radiologists, so there are international differences in how we estimate the density as radiologists. So if the software is sort of is trained on certain radiologists, it's something to take into account that also could affect their agreement. So what's the potential use of the VDS for mammography and Thomas synthesis in clinical practice? And that's one of the most important things, I think. So today most screening programs looks like this, so it's the same model for everyone. The same screening program, and in Sweden we're just the screening program. It's likely by age, with younger women being screen data slide smaller interval that is 18 months. In contrast to women between 50 and 74 being screen every 24th month. I think we would like it to screening program to be a bit more like this so personalized or individualized an I think density is a great tool that can help us here and there's of course of course a lot of interesting trials going on here. I'm just going to make a few examples. So density could help us as a factor in creating risk scores and the Swedish studies cited at Kolinski instituted. The Kalama study. Are doing that, taking in density into account creating risk scores and also the British Progress study. Density in primary prevention and also an offspring from the Swedish Chemist study is the Charisma study, also including collaborators from our research group. They are at present including patients and they looking at the smallest amount of tamoxifen to reduce density in the primary prevention setting. So it's going to be very interesting to follow that and they also combining it with research on the preclinical level. To see what actually happens in the breast before and after tamoxifen regarding the density. Density of course, as a tool in individualizing screening programs, and we got the very exciting wisdom trial in the US and looking. It's a randomized controlled trial and they're also looking at personalized screening, taking density into account and the Dutch dense trial where women with dense breast get screen with a supplemental MRI. And finally density to monitor treatment response. Another study called the Neo dense trial, cited at the University Hospital where I work, also including our research group. Are monitoring a treatment monitoring density as a proxy for treatment response in the neoadjuvant setting in women with breast cancer and that studies also previously including patients. So finally some takehome messages. Volumetric breast density analysis is a proven tool to estimate density in mammography and Thomas synthesis. Of course, larger studies are needed, I think especially for tomosynthesis. Differences towards radiologists may be noted, but not an obstacle. I mean, the focus again should be on which method the best can show. The Association between density and risk, and perhaps also the method that works with your particular workflow. And finally, breast density may help us towards personalized imaging and care. I would like to thank my funding and our research group Lind University Cancer Imaging Group and thank you all for listening.

SIEMENS Healthineers Cover ECR 2017 Reliability of Automated Breast Breast Density Assessment Using Breast Access OF Tomosynthesis Images Breast Cancer Risk and Volumetric Breast Density Estimation from Full-Field Erg al. &east Caret Res&rch 2014. 16:439 Erg et al. &east Caret Re-s&rch 2014. 16:439 era' Breast 2014. Breast Cancer V BDA vs. radiologists VBDA vs. radiologists Learning objectives Density Measurementsl Breast density measurements Radiologist vs. software Radiologists vs. software Take home messages (DBT) "'5/439 LUND Digital Mammograms: A Validation Study RESEARCH UNIVERSITY Mammographic Density Albert Bram M. Robert M. Robert M. Pontus Andreas Fieselmann2, Magnus Pontus Andreas Fieselmann2, Magnus Petersson , Pontus Andreas , Magnus Volumetric breast density Volumetric breast density Combined Hanna Sartor* , Kristina Daniel Förnvikl, and Sophia Zackrisson Hanna Sartor3, Kristina Daniel and Sophia Zackrisson Hanna Sartor* , Hanna Sartor* , Kristina Daniel and Sophia Zackrisson Hanna Sartor* , Kristina Daniel , and Sophia Zackrisson Hanna Sartor* , Kristina Daniel , and Sophia Hanna Sartor* , Kristina Daniel and Sophia Hanna Sartor* , Kristina Daniel Förnvikl, and Sophia Hanna Kristina Daniel , and Sophia Hanna Kristina Daniel and Sophia Volumetric breast density inalyses Assessed with Semiautomated RESEARCH ARTICLE Open Access The Breast Quantitative Population To Results Differences in breast density between modalities Qualtitative Quantitative Medical Radiation Physics. Departnrnt of Translational Medicine, Lund University, • To understand the clinical basics of • Volumetric breast density analysis is a Volumetric breast density Combined To understand the clinical basics of Volumetric breast density analysis is a mammography and tomosynthesis — and Fully Automated Methods homepage: homepage: c homepage: co Automated technques for measuring breast density snow good Automated techniques for measuring breast density show good Digital mammographic density and breast cancer Digital DOI: DOI Morrish et al Volumetric breast density volumetric breast density analyses Volumetric breast density analyses Volumetric breast density an proven tool to estimate density in SIEMENS Radiology 2015 Radiology 2016 correlation, but these are poorly correlated with observer's pontus . timbergamed. Iu. se pontus . timbergamed. Iu. pontus . timbergamed. I u . pontus . timber-gamed. Iu . se pontus . timbergamed. I u . se pontus . timber-gamed. I u. se pontus . timber-gamed. Iu. pontus . timbergamed. Iu . pontus . timbergßmed. Iu. Agreement of Mammographic Measures of Volumetric . Wolfe (MRI, FFDM, DBT) l. Hanna Sartor, MD PhD • Area-based Hanna Sartor, MD PhD I. Hanna Sartor, MD PhD M. article and BI-RADSI Sienrns Healthineers. Erlangen. Germany Sienrns Healthineers, Erlangen, Germany Sienrns Healthineers, Erlangen. Germany risk: a case—control study of six alternative density risk: a case-control study of six alternative density BREAST Breast Density to MRI To acknowledge the difference between To discuss VBDA's potential use for (VBDA) based on previous studies mammography and tomosynthesis mammography and digital breast Consistency Of breast density categories in serial screening Skåne University Hospit01 Skåne University Hospital Malmnö Medical Radiation Physics. Departnrnt of Translational Medicine, Lund University, European Radiology Medical Radioliog Physics. Department of Translational Medicine. Lund University, Medical Radioliog Phy,ics. Department of Translational Medicirr, Lund University, 3 Medical Radiation Physics. Department of Translational Medicine, Lund University, Medical Radiation Physics. Departnrnt of Translational Medicine. Lund University, Medical Radiology. Department of Translational Medicine. Lund University, 3 Medical Radiation Physics. Departrnrnt of Translational Medicine, Lund University, Medical Radiation Physics, Departnrnt of Translational Medicine, Lund University, Skåne University Hospital Malmö . Tabar — Machida et al Springer Plus 2016, Pertuz et al Radiology — Machida et al Springer Plus 2016, Pettuz et al Radiology — Cumulus, ImageJ, Densitas mammograms: A comparison between automated and human Jeff Ania Jeff Ania Bo Ania Skåne University Hospital, 205 02 Malmo, Sweden Skåne University Hospital. 205 02 Malmo. Sweden Skåne University Hospital. 205 02 Malmo, Sweden assessment methods Malmö, Sweden mammography and tomosynthesis radiologists' assessment of breast mammography and digital breast Nola A. artor et al There was moderate agreement (K: 0.55 (0.53-0.56)) between Abra M. 2015, Tagliafico et al Breast Cancer Res Treat 2013 Unilabs AB, 205 02 Malmo. Sweden Unilabs AB. 205 02 Malmo. Sweden Unilabs AB. 205 02 Malmo, Sweden Unilabs AB, 205 02 Malmo, Sweden Healthineers assessment Siemens Healthcare GmbH. Erlangen, Germany Sienrns Healthineers, Erlangen. Germany Purpose: purpose: To comparv• three metrics of breast density on full-field TO compan• three metrics of breast density on compare three metrics of breast density on fulVfieId Sartor et al • BI-RADS • Volumetric (automated) Measuring mammographic density: comparing a fully automated European Radiology volumetric density and BI-RADS scores from radiologists digital mammographic (FFDM) images as predictors of density and software measurements To acknowledge the difference between tomosynthesis (DBT) in clinical practice Differences towards radiologists may be O Springer International Publishing Switzerland 2016 European Katharina Holland Jan van Zelst Gerard J. den Heeten Mechli Imhof-Tas Katharina Holland Jan van Zelst Gerard J. den Heeten Mechli Imhof-Tas a. Jah A. Amarda Erg , Galanti t, Jchn Valerie McCormack4, Jingrrei Lis, Mtch Dowsett6, Sarah Amarda Erg , Galanti t, Jchn Sheptuæ, Valerie McCormack4, Jingrrei Lis, Mtch Dowsett6, Sarah future breast cancer risk. breast risk. Ritse M. Mann Carla H. van Gils d. Nico Karssemeijer Ritse M. Mann Carla H. van Cils d. Nico Karssemeijer volumetric assessment versus European radiologists' qualitative H. Radiology 2015 A. Tingberg al. (Eds.): IWDM 2016, LNCS 9699, 197-202, 2016. A. Tingberg al. (Eds.): IWDM 2016, LNCS 9699. pp. 197-202, 2016. A. Tingberg et al. (Eds.): IWDM 2016, L-NCS 9699, 197-202, 2016. A. Tingberg al. (Eds.): IWDM 2016, LNCS 9699, pp. 197-202, 2016. A. Tingberg et al. (Eds.): IWDM 2016, I-NCS 9699, pp. 2016. A. al. (Eds.): IWDM 2016, LNCS 9699, pp. 197-202, 2016. A. Tingberg et al. (Eds.): IWDM 2016, 9699, pp. 2016. H, noted but not an obstacle radiologists' assessment of breast Steve Allen" and Isabd dos-Santos-Silva'" — Quantra, VBDA (Siemens), Volpara, Volumetric breast density Volumetric breast density an Volumetric breast density in Volumetric breast density analyses Volumetric density in Volumetric breast density Combined volumetric breast density analyses Materials and Materials institutional board-approved study included institutional board—approved study included DOI: DOI: 26 DOI: 10.10071978-3-319-41546-8_26 DOI: 10.1007/978-3-319-41546-8_26 Methods: 125 women with invasive cancer and 274 ago- and 125 women with invasive breast cancer and 274 ago- and 125 women with invasivo cancer and 274 ago, and 125 women with invasive breast cancer and 274 and classification radiologists analysis in mammography and density and software measurements race-matched control subjects who screening an det Waal et al SXA Volumetric measures obtained using the Volpara or Quantra algorithms "Volumetric measurements obtained from FFDM (Volpara) show high tomosynthesis — brief overview PLOS ONE 2015 Risk scores have excellent reliability for repeated measures, while the Cumulus Volumetric breast density Volumetric density in Volumetric breast density an Hanna Sartor • Kristina Ling • Aldana Rosy)' • Signe Borgquist Hanna Sartor • Kristina Ling • Aldana ROSY)' • Signe Borgquist Hanna Sartor • Kristina Ling • Aldana • Signe Borgquist correlation with MRI data" data Sophia Zackrisson • Pontus Timberg• Sophia Zackrisson • Pontus Timberg' To discuss VBDA's potential use for measures have moderate reliability" Breast 2016 mammography vs tomosynthesis (central projection): 186 exams from women with high familial risk of breast cancer Primary prevention Holland et al 30 women with repeated imaging by two technologists high correlation (r = 0.94) substantial agreement (K = 0 76) high correlation (r = 0.94) substantial agreement (K = O. 76) high correlation (r = 0.94) substantial agreement (K = 0. 76) mammography and digital breast mammography and tomosynthesis UNIVERSITY Breast 2016 he Breast 2016 • Masking vs. volume Screening programs tomosynthesis (DBT) in clinical practice Substantial agreement between radiologists (BI-RADS 5th Ed) and "The categorization of serial mammograms in density classes is more "Moderate agreement (K: 0.55 (0.53-0.56)) between consistent with automated software (Volpara) than with a mixed group of software volumetric density and BI-RADS 4th Ed scores from radiologists" • Processed vs. raw data Monitor treatment response • mammography images (K = 73) tomosynthesis images (K mammography images (K = 0. 73) tomosynthesis images (K mammography images (K = O. 73) tomosynthesis images (K mammography images (K = 0. 73) tomosynthesis images (K 62) mammography images (K = O. 73) tomosynthesis images (K 62) human readers" 500 pair of subsequent screening (1000 exams) LUND • International differences between radiologists? Wolfe et al 1976; Gram et al 1997, D'Orsi et al 2013, Byng et al 1994; Densitas.ca; UNIVERSITY Shepard et al 2011; Ciatto et al 2012 Fieselmann et al 2016; Highnam et al 2010,