Superior Diagnostic Accuracy with Digital Breast Tomosynthesis plus Synthetic Mammograms White Paper

White paper Superior Diagnostic Accuracy with Digital Breast Tomosynthesis plus Synthetic Mammograms Key findings of the clinical Premarket Approval study on the MAMMOMAT Revelation with Tomosynthesis Option siemens-healthineers.com/tomosynthesis Daan Hellingman, PhD Clinical Excellence & Cooperations, Siemens Healthcare GmbH Nancy A. Obuchowski, PhD Quantitative Health Science, Cleveland Clinic Foundation Agnieszka Lazar Research and Development, Siemens Healthcare GmbH Mathias D. Hoernig X-ray Physics and Systems, Siemens Healthcare GmbH Axel Hebecker, PhD Clinical Excellence & Cooperations, Siemens Healthcare GmbH SIEMENS Healthineers White paper · MAMMOMAT Revelation with Tomosynthesis Option Introduction The aim of this white paper is to summarize the results of a multi- reader multi-case (MRMC) study on the MAMMOMAT Revelation with Tomosynthesis Option that laid the groundwork for its Premarket Approval (PMA) by the Food and Drug Administration (FDA) [1]. Part of these results contributed to the assessment of the safety and effectiveness of 2-view 50° wide-angle digital breast tomosynthesis (DBT) plus synthetic mammograms for breast cancer screening and diagnosis in the United States. 2 MAMMOMAT Revelation with Tomosynthesis Option · White paper Rationale, objective and study design Siemens Healthineers has been the first vendor that Insight 2D is reconstructed from a DBT data set using received FDA approval for the use of 2-view DBT alone special algorithms. The soft tissue contrast in the in breast cancer screening [2]. Although 2-view wide- synthetic mammograms has been optimized so that angle DBT has been proven to have a superior diagnostic the overall image impression of Insight 2D1 is very accuracy as compared to 2-view full-field digital similar to FFDM (see Figure 1). mammography (FFDM) alone [3,4], additional 2D mammograms are often acquired in clinical practice The primary objective of this MRMC reader study was that serve as overview images and enable comparison to demonstrate the superiority of 50° wide-angle DBT with prior 2D mammograms. Unfortunately, combining plus synthetic mammograms in terms of the diagnostic DBT and FFDM results in a longer acquisition and breast accuracy in detecting and characterizing breast lesions compression time for the patient. More importantly, as compared to FFDM alone. FFDM as an adjunct to DBT generates a higher radiation dose. To overcome these challenges, the so-called 2D synthetic mammogram (Insight 2D) was developed. a b c d Insight 2D Study ID 1aab084 Study ID 1aab096 Study ID 1aab070 Study ID 1aab077 FFDM Figure 1: Comparison of Insight 2D and FFDM for different breast densities 1 Software version VC20 or higher 3 White paper · MAMMOMAT Revelation with Tomosynthesis Option Library of images MRMC study population This study is based on a library of images specifically A subset of cases was randomly selected from the library collected for this and previous DBT PMA studies. A total of images (Table 1). Specific mammographic features like of 764 subjects (54.9 ± 9.8 years) were enrolled between breast density and lesion type were included in order to May 2011 and April 2014 from seven United States obtain a distribution of cases similar to that seen in a clinical sites: screening population. • Duke University, Durham, NC; Table 1: Subject characteristics • SUNY, Stonybrook, NY; • Brigham & Women’s Hospital, Boston, MA; PICKS • Cleveland Clinic, Cleveland, OH; vs. • NYU, New York, NY; • St. Luke’s Episcopal Hospital, Houston, TX; and 2-view DBT + Insight 2D 2-view FFDM • Miami Baptist Hospital, Miami, FL. The library included FFDM images and DBT images. Subjects selected: 350 All images were collected prospectively through a • 111 with cancer written informed consent process with approval from • 90 with benign findings1 local institutional review boards prior to enrolling • participants in image collection. 149 normal1 The FFDM images were acquired according to the Breast density standard of care using a number of commercially • Almost entirely fat: 6.9% available FFDM systems. In addition to these clinically • indicated screening mammograms, the DBT images Scattered fibroglandular: 38.4% were acquired solely on Siemens Healthineers systems. • Heterogeneously dense: 47.1% All breasts were imaged in craniocaudal (CC) as well as • Extremely dense: 7.5% mediolateral oblique (MLO) positioning. • Missing data: 1.1% For malignant lesions, the ground truth for the type and location of the lesions was 1) based on the Malignant lesion type mammography findings described by the radiologist at • Mass: 64.8% the clinical site and based on the standard of care and • Calcification: 21.1% 2) confirmed by the radiology and pathology reports after biopsy. For cases with a biopsy-confirmed benign • Architectural distortion: 9.2% finding, follow-up examinations were performed after • Asymmetric density: 4.9% 6 or 12 months. For normal findings, follow-up was performed after 12 months to confirm the non-cancer Cancer type status. • Invasive cancer: 81% • Ductal carcinoma in situ: 19% 1 23 breasts in 19 subjects (10 with benign and 9 with normal findings) had unknown ground-truth due to missing follow-up or BI-RADS 3 at follow-up imaging. Subjects were included to prevent selection bias. Statistical analysis shows that the results with imputation of the unverified breasts are consistent with the results based on only verified breasts. 4 MAMMOMAT Revelation with Tomosynthesis Option · White paper Reading and image interpretation A total of 20 Mammography Quality Standards Act (MQSA) – qualified radiologists (“readers”) reviewed and scored the clinical images. All readers were trained to read Siemens Healthineers wide-angle DBT + synthetic mammograms. Each reader participated in two reading sessions separated by a wash-out period of at least 6 weeks. During each reading session, half of the cases were interpreted using FFDM alone (MLO and CC). The other half of the cases were interpreted using DBT (CC and MLO) + Insight 2D (CC and MLO). After the wash-out period, the readers interpreted the opposite modality for each case. Cases were randomized to be read with either FFDM or DBT + Insight 2D first and the reading order within each group was randomly assigned. Readers assigned a screening BI-RADS score to each breast (0, 1, or 2). Readers were also asked to identify all suspicious lesions and assign a probability of malignancy score to each. 5 White paper · MAMMOMAT Revelation with Tomosynthesis Option MRMC study results Significant improvement in 1.00 diagnostic accuracy Table 2 summarizes the main results that were observed 0.75 in terms of diagnostic accuracy. To calculate sensitivity and specificity, a BI-RADS score of 1 or 2 was considered a negative finding, and a BI-RADS score of 0 was considered a positive finding. 0.50 The diagnostic accuracy in detecting and characterizing breast lesions was described by referring to the area Mean sensitivity under the receiver operating characteristic curve (AUC ROC). The probability of malignancy score was used to 0.25 construct the ROC curves. The readers’ average area Modality under the ROC curve was calculated to summarize the FFDM diagnostic accuracy of all readers together. The average DBT + Insight 2D AUC ROC was significantly higher (+ 0.056) with DBT + 0.00 Insight 2D compared to FFDM alone (Figure 2). Nineteen 0.00 0.25 0.50 0.75 1.00 readers (95%) improved their accuracy in detecting and FPR diagnosing cancers when reading 50° wide-angle DBT + Figure 2: Average reader breast-level ROC curves for 2-view FFDM Insight 2D compared to FFDM alone (Figure 3). (orange) and 2-view DBT + Insight 2D (petrol). Calculation based on ratings of 20 readers. These results demonstrate the superior diagnostic accuracy of Siemens Healthineers’ 50° wide-angle DBT 0.95 plus Insight 2D compared to FFDM alone. 0.9 Table 2: Significant improvement in diagnostic accuracy 0.85 0.8 elseloks vs. 0.75 2-view DBT + Insight 2D 2-view FFDM 0.7 Estimated AUC Average AUC ROC: 0.65 • 2-view FFDM: 0.837 0.6 • 2-view DBT + Insight 2D: 0.893 ABCDEFGHIJKLMNOPQRST • ∆AUC ROC: 0.056 Reader ID Figure 3: Readers‘ estimated breast-level AUCs with FFDM sorted by • p-value: < 0.0001 lowest to highest (orange) and the estimated increase (petrol) or decrease (grey) with DBT + Insight 2D. 6 MAMMOMAT Revelation with Tomosynthesis Option · White paper Significant improvement in sensitivity Significant reduction in the non-cancer recall rate The sensitivity in detecting breast cancers was calculated at the breast-level. Among breasts with cancer, a BI-RADS All women with a BI-RADS 0 during screening will be score of 0 assigned to the breast was considered a true- recalled for further diagnostic work-up. This group positive finding. consists of true and false positives, the latter receiving The readers’ average sensitivity was significantly higher unnecessary follow-up procedures like ultrasound or with DBT + Insight 2D compared to FFDM alone (Table 3). biopsy. The non-cancer recall rate was calculated at the The average relative increase in sensitivity at the subject- subject-level and describes the number of women with level was 5.8%. Eighteen readers (90%) improved their false positive findings as a proportion of all women sensitivity when reading 50° wide-angle tomosynthesis without cancer. with Insight 2D compared to FFDM alone. The readers’ average non-cancer recall rate was signifi- This demonstrates the superior sensitivity of cantly lower with DBT + Insight 2D compared to FFDM Siemens Healthineers’ 50° wide-angle DBT plus Insight alone (Table 4). The average relative reduction in non- 2D compared to FFDM alone. cancer recall rate at the subject-level was 27.4%. Nine- teen readers (95%) reduced their non-cancer recall rate when reading 50° wide-angle tomosynthesis with Insight 2D compared to FFDM alone. This demonstrates the increased diagnostic confidence obtained using Siemens Healthineers’ 50° wide-angle DBT plus Insight 2D compared to FFDM alone. Table 3: Significant improvement in sensitivity Table 4: Significant reduction in the non-cancer recall rate vs. vs. 2-view DBT + Insight 2D 2-view FFDM 2-view DBT + Insight 2D 2-view FFDM Readers’ average sensitivity: Readers’ average non-cancer recall rate: • 2-view FFDM: 0.8047 • 2-view FFDM: 0.4729 • 2-view DBT + Insight 2D: 0.8517 • 2-view DBT + Insight 2D: 0.3435 • ∆sensitivity: 0.047 • ∆Non-cancer recall rate: -0.1294 • p-value: 0.0430 • p-value: < 0.0001 • Relative increase: 5.8% • Relative decrease: 27.4% 7 White paper · MAMMOMAT Revelation with Tomosynthesis Option Trends in specific breast cancer subgroups Supplemental analyses were performed to observe trends in different breast cancer subgroups. Table 5 summarizes the differences in average reader breast-level AUCs, as measure of diagnostic accuracy, between wide-angle DBT plus Insight 2D and FFDM. Table 5: Average reader breast-level AUCs for specific breast cancer subgroups DBT + Insight 2D (STD) FFDM (STD) ∆AUC Dense breasts (c/d) 0.8754 (0.0229) 0.8295 (0.0255) + 0.0459 Non-dense breasts (a/b) 0.9088 (0.0208) 0.8391 (0.0259) + 0.0697 Mass 0.9363 (0.0145) 0.8714 (0.0198) + 0.0649 Calcification 0.8094 (0.0367) 0.8159 (0.0360) 0.0064 - Architectural distortion 0.9298 (0.0313) 0.8414 (0.0421) + 0.0884 Invasive cancer 0.9273 (0.0147) 0.8535 (0.0120) + 0.0739 8 MAMMOMAT Revelation with Tomosynthesis Option · White paper Summary This white paper describes a clinical study using Siemens Healthineers’ wide-angle DBT system plus synthetic mammograms (Insight 2D) which supported the Premarket Approval by the Food and Drug Adminis- tration. The MRMC study demonstrated a superior diagnostic accuracy, sensitivity, and a reduced non- cancer recall rate using Siemens Healthineers’ 50° wide-angle DBT plus Insight 2D as compared to FFDM alone. These results provide evidence that Siemens Healthineers’ 50° wide-angle DBT plus Insight 2D can be used with a high level of confidence in breast cancer screening and diagnosis. Abbreviations References AUC ROC Area under the ROC curve [1] Siemens Medical Solutions USA Inc. (2020) BI-RADS Breast Imaging Reporting and PMA P140011/S007 Data System [2] Siemens Medical Solutions USA Inc. (2016) CC Craniocaudal PMA P140011/S001 Digital Breast Tomosynthesis [3] Georgian-Smith D, Obuchowski NA, Lo JY, Brem RF, DBT Baker JA, Fisher PR, Rim A, Zhao W, Fajardo LL, FFDM Full-Field Digital Mammography Mertelmeier T. Can Digital Breast Tomosynthesis MLO Mediolateral oblique Replace Full-Field Digital Mammography? A Multireader, Multicase Study of Wide-Angle MRMC Multi-Reader Multi-Case Tomosynthesis. AJR Am J Roentgenol. 2019:1-7 MQSA Mammography Quality Standards [4] Nalleweg N, Mertelmeier T, Korporaal JG, Act and Program Hebecker A (2016) Superior Diagnostic Accuracy PMA Premarket Approval with Additional Digital Breast Tomosynthesis: Main findings of the clinical study for Premarket ROC Receiver Operating Characteristic Approval of Mammomat Inspiration with STD Standard Deviation Tomosynthesis Option. Siemens White Paper 9 On account of certain regional limitations of sales rights The statements by Siemens’ Healthineers customers and service availability, we cannot guarantee that all described herein are based on results that were achieved products included in this brochure are available through in the customer’s unique setting. Because there is no the Siemens Healthineers sales organization worldwide. “typical” hospital or laboratory and many variables exist (e.g., hospital size, samples mix, case mix, level of IT Availability and packaging may vary by country and and/or automation adoption) there can be no guarantee are subject to change without prior notice. Some/All of that other customers will achieve the same results. the features and products described herein may not be available in the United States. The information in this document contains general technical descriptions of specifications and options as well as standard and optional features that do not always have to be present in individual cases. Siemens Healthineers reserves the right to modify the design, packaging, specifications, and options described herein without prior notice. Please contact your local Siemens Healthineers sales representative for the most current information. Note: Any technical data contained in this document may vary within defined tolerances. Original images always lose a certain amount of detail when reproduced. Siemens Healthineers Headquarters Siemens Healthcare GmbH Henkestr. 127 91052 Erlangen, Germany Phone: +49 9131 84-0 siemens-healthineers.com Published by Siemens Healthcare GmbH · 11104 1021 online · ©Siemens Healthcare GmbH, 2021