In work packages WP1-3, we designed and built a DBTMI prototype system by combining a clinical DBT system (Mammomat Inspiration, Siemens, Erlangen) with a commercial pressure sensor (BRE 5350, Tekscan, Boston) placed on the breast support of the DBT system. An acquisition protocol was developed for sensor positioning and the x-ray parameters selection. After imaging a breast, we acquire a DBT image of a uniform plastic (Lucite) block on top of the sensor. Lucite images are used for preprocessing to reduce artifacts in DBT images, caused by the presence of the sensor during simultaneous DBTMI acquisition. Reconstruction of DBT images (WP3) has been developed in collaboration with our industrial partner (Real Time Tomography, Villanova). The prototype has been used for the phantom and clinical evaluation in WP4-5.
WP4-5 focused on the evaluation and optimization of the DBT prototype, both preclinically and in pilot clinical tests. Preclinical evaluation (WP4) has used simulations/VCTs and physical breast phantoms. For VCTs of DBTMI, we developed a method to simulate the MI acquisition. We also developed methods to select the population of virtual patients, and a model of tumor growth, to simulate multiple years of breast screening. Simulation work was performed by Master students mentored by Dr Bakic and published in a journal paper (2 more in review) and several conference papers. Phantom evaluation has used radiographic undeformable phantoms and a multimodality deformable breast phantom. Undeformable phantoms were used to analyze sensor artifacts. Deformable phantoms were used to test the MI acquisition and composite images of MI and DBT data. Results from phantom studies were presented in various conference papers and grant funding applications. An example of a DBTMI composite image of a deformable phantom is shown in Project Illustration.
Clinical tests of DBTMI (WP5) have been performed at the SUS Breast Clinic in Malmö since March 2021, approved by Swedish Ethical Review Authority. The collection of clinical data was delayed due to the COVID-19 pandemics. After the end of the fellowship period, we have continued with the clinical collection (in agreement with our SUS collaborators). To date we collected DBTMI datasets from 75 women. The goal is to complete the proposed 150 acquisitions. The preliminary analysis of clinical data is ongoing. Sample clinical DBTMI images are shown in Project Illustration; they were also used in our grant applications. Clinical tests elucidated important aspects, e.g. DBTMI scheduling due to the need to acquire Lucite images. The exploratory integration of artificial intelligence (AI) in DBTMI (WP6) is ongoing. Initial analysis suggested a significant correlation of clinically used AI scores and our MI data for biopsy-confirmed cancers; other findings were not correlated. Further tests will be performed upon the completion of clinical DBTMI data collection.
In addition to research aims, this MSCA fellowship has fulfilled the professional development objectives. During his visit, Dr Bakic supervised 3 Master students (plus 1 Master and 1 PhD student ongoing), taught 3 courses at Lund and 1 course at a US medical imaging conference (highly rated), published 6 journal and 19 conference papers (3 journal papers pending), submitted 12 grant proposals of which 5 funded (plus 3 pending). After the fellowship, Dr Bakic has continued employment at Lund Diagnostic Radiology as a Researcher, while applying for a faculty position.
Our results have been broadly exploited and disseminated in 15 proffered and 2 invited talks at scientific conferences, 7 seminars, and 4 news reports. Notable presentations include Dr Bakic’s online talks at 2020 Workshop on Breast Imaging and 2020 Meeting on Optimisation in X-ray and Molecular Imaging; and talks by Dr Bakic’s students: Ms Axelsson at 2021 SPIE Medical Imaging Conference and Ms Tomic at 2020 Food and Drug Administration Seminar and 2021 European Congress of Radiology.
In conclusion, this fellowship has produced a DBTMI prototype, versatile tools for its evaluation, and a preliminary set of clinical data. The research achievements, with rich and diverse professional development, supported Dr Bakic in his reintegration into European research community.
