Program Information
Deformable Mapping Technique to Correlate Lesions in X-Ray and Ultrasound Breast Images
C Green1*, M Goodsitt1 , K Brock2 , C Davis3 , P Carson1 , E Christodoulou1 , E Larson1 , (1) University of Michigan, Ann Arbor, MI, (2) UT MD Anderson Cancer Center, Houston, TX, (3) General Electric CRD, Niskayuna, NY
Presentations
WE-G-601-8 (Wednesday, August 2, 2017) 4:30 PM - 6:00 PM Room: 601
Purpose: To develop a deformable mapping method to match corresponding lesions in dedicated Breast CT (DBCT), automated breast ultrasound (ABUS), and digital breast tomosynthesis (DBT) images.
Methods: A CIRS multi-modality breast phantom containing 10 lesions was employed and imaged with conventional CT to simulate coronal slice acquisition for DBCT (without compression), DBT (upright positioning with cranial-caudal compression), and ABUS (supine positioning with anterior-to-chest wall compression). The phantom images were segmented manually and finite element models (FEM) were generated using OptiStruct FEM solver and Morfeus to deformably map and match lesions across imaging modalities. Performance was assessed with measures of center of mass (COM) displacements for all corresponding lesions and minimum edge distances between corresponding lesions that did not overlap.
Results: For mapping of DBT to ABUS, 7 of 10 lesions were correlated with a mean COM displacement of 1.06 ± 0.65 cm and a mean minimum edge distance of 0.64 ± 0.30 cm. For mapping of DBCT to DBT, 9 of 10 lesions were correlated with a mean COM displacement of 0.82 ± 0.34 cm and a mean minimum edge distance of 0.19 ± 0.00 cm. Corresponding results for mapping of DBCT to ABUS, included 8 of 10 correlated lesions with a mean COM displacement of 2.05 ± 1.10 cm and a mean minimum edge distance of 0.86 ± 0.07 cm. Of the three mapping cases studied the largest COM displacement was 3.61 cm for a lesion in DBCT to ABUS mapping; however, the corresponding minimum edge displacement was only 0.80 cm.
Conclusion: This work demonstrates the potential for a deformable mapping technique to relate lesions within various breast imaging modalities. This should improve radiologists’ characterization of breast lesions across modalities, reduce patient callbacks and unnecessary biopsies. Future work will include automated segmentation and optimized deformable finite element modeling.
Funding Support, Disclosures, and Conflict of Interest: Supported in part by a grant from GE Global Research (15-PAF04328)
Contact Email: