Simultaneous digital breast tomosynthesis (DBT) and mechanical imaging (MI) – DBTMI
Project description: DBTMI is motivated by a challenge of underdiagnosis and false positives (FPs) in breast cancer healthcare. Mammography screening reduces mortality by early cancer detection. It is limited by FPs and unnecessary biopsies, with related complications, anxiety, and burden on the healthcare system. MI maps local pressure on the breast compressed during x-ray exam. Increased pressure due to cancer can resolve suspicious x-ray findings - and reduce recalls. Simultaneous imaging prevents increased exam time or radiation dose.
In this project, we have combined the extensive experience of LU in MI and clinical imaging, and Predrag’s expertise in VCT and imaging system evaluation, enabling significant benefits. Within the two-year timeline, we have designed and built a DBTMI prototype system, and developed image processing and reconstruction to maximize image quality. We have been evaluating the DBTMI prototype, first preclinically by VCTs and physical phantoms, followed by a pilot clinical trial. Analysis of clinical data is ongoing, as well as the exploration of modern AI methods to improve DBTMI performance.
Implications for healthcare: The most exciting recent innovations in breast cancer imaging include: DBT (increased cancer detection), MI (reduced FPs and unnecessary biopsies), AI (efficient analysis of clinical images) and VCTs (efficient preclinical evaluation of imaging systems). Our project integrates these innovations to improve clinical accuracy, while preserving the clinical workflow. The expected benefits are multifaceted: improved screening program, tools for clinical decision making (using both radiographic and MI information), and affordable and sustainable upgrade of breast cancer care.
Implications from patient perspective: DBTMI offers a win-win opportunity to achieve better clinical accuracy while preserving the patent experience – without increased exam time and radiation dose. Reducing FPs prevents many unnecessary biopsies, anxiety, absence from work and financial cost. We have been collaborating with clinical and qualitative research specialists to adequately inform women about DBTMI and incorporate their response and suggestions into our project.
Implications for research: The optimized DBTMI prototype supports research toward the use of MI in image-guided breast interventions, neoadjuvant follow-up, cancer risk evaluation – and potentially other applications beyond breast cancer. The unique VCT approach have been expanded to add the simulation of MI and ultrasound imaging. These substantial innovations provoke new ideas, foster collaborations, and most importantly push clinical cancer care forward!
Major achievements to date:
- Built a DBTMI prototype by combining a clinical DBT system with a commercial pressure sensor
- Developed preprocessing and reconstruction of DBT images in DBTMI, to reduce sensor artifacts;
- Developed simulation tools for preclinical evaluation of the prototype using VCTs
- Performed various DBTMI tests using breast phantoms, undeformable and deformable
- Started pilot collection of clinical DBTMI data with ethics approval; 71 cases collected to date. Clinical collection is ongoing.
- Preliminary analysis of DBT vs MI (from clinical DM+MI data) suggests ~20% improvement with DM vs. DBT only. Analysis of clinical DBTMI data is ongoing.
- Preliminary analysis of previous clinical data observed significant correlation between AI scores from DM and corresponding MI data for biopsy confirmed cancers; other findings were not correlated. Integration of AI with DBTMI data is pending.
- Initiated development of physical phantoms for MI using 3D printers
- Mentored 3 Master students; currently mentoring 1 Master and 1 PhD student
- Taught 3 courses at Lund and 1 VCT course at 2020 SPIE Medical Imaging conference (highly rated). A CME/ST course at Annual Meeting of the Swedish Association for Radiation Physics scheduled fr Spring 2022
- Published 1 journal paper (with 5 other related and 3 pending) ad 14 conference papers (with 5 related).
- Submitted 12 grant proposals – of which 5 funded (with 3 pending)
- Established several research contacts, resulting in joint grant applications. Motivated research in ethics and nursing of breast screening. Participated in outreach activities by Bröstcancerfårbundet
Penn: B Barufaldi, A Maidment, E Conant
Others: S Ng (Real Time Tomography), O Diaz (Univ Barcelona), I Sechopoulos, J Teuwen (Radboud Univ), H Bosmans (Katholieke Univ Leuven), J Wicklein, S Kappler (Siemens Healthcare), T Gustafsson, C Malacaria (Tekscan, Inc.), A. Rodriguez Ruiz, N Karssemeijer (ScreenPoint Medical), A Bjällmark (Jönköping Univ)
“This project has received funding from the European Union’s Horizon 2020 research and
innovation programme under the Marie Skłodowska-Curie grant agreement No 846540”.
Additional funding provided by Bröstcancerförbundet, Allmänna Sjukhusets i Malmö Stiftelse för bekämpande av cancer, and Malmö Cancer Center
- Preclinical optimization of DBTMI prototype: We built a DBTMI prototype by combining a clinical DBT system with a commercially available MI sensor. To achieve simultaneous DBTMI, we need to reduce artifacts visible in DBT images acquired with MI sensor present. To that end we have developed a preprocessing of DBT projections, which combines x-ray images of the breast with the MI sensor with the sensor images without the breast. We are evaluating and optimizing the preprocessing to ensure diagnostic quality of DBT images acquired during DBTMI. This preclinical evaluation is performed by imaging physical breast-like phantoms and with computer simulations by VCTs.
Publications: Barufaldi (Radiation Protection Densitometry, 2021); Bakic (SPIE 2019, SPIE 2020, IWBI 2020); Axelsson (SPIE 2021); Tomic (SPIE 2021, ECR 2021);
Funding: EU H2020 Marie Curie fellowship; BF
- Acquire clinical DBTMI data: Recently (April 2021) we received Ethics approval and are about to start collecting clinical DBTMI data. The pilot collection of patient DBTMI data should confirm the clinical image quality and optimize the clinical protocol (including the acquisition of senor images without the breast).
Funding: EU H2020 Marie Curie fellowship, CF (Zackrisson)
- Analyze DBTMI data: Analysis of simulated and clinical DBTMI data will be focused on (i) the potential classification of different molecular tumor types and identification of aggressive cancers, and (ii) optimization of clinical image display and review of composite DBTMI data.
Funding: Pending (VR, US DOD, CF)
- Explore use of AI in DBTMI optimization (lead: M. Dustler) The ongoing AI revolution in clinical imaging has motivated us to explore the use of deep learning networks for additional optimization of DBTMI. We see potential for AI in optimizing (or even replacing) DBTMI preprocessing step, to automatic selection of patient with max benefits by DBTMI screening, to AI based analysis and identification of aggressive tumors. An exploratory work may be performed upon collection of pilot clinical DBTMI datasets.
Publications: Bejnö, IWBI 2020
Funding: EU H2020 Marie Curie fellowship exploratory aim
- Support clinical adoption of DBTMI: The pilot acquisition should be followed by collection of clinical DBTMI and DBT+US data (funding and Ethics pending), to compare their clinical performance and encourage clinical adoption of DBTMI.
Funding: Pending (VR, US DOD, CF)
- (future) Explore future DBTMI applications: Our screening DBTMI research has motivated forward looking ideas on potential applications of DBTMI to guide biopsies or improve the estimation of breast density and cancer risk.