Premium
TU‐CD‐207‐07: Creating a Synthetic Mammogram From a Digital Breast Tomosynthesis Image
Author(s) -
Hong J,
Lee T,
Kim H,
Choi Y,
Lee J,
Cho S
Publication year - 2015
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.4925626
Subject(s) - artificial intelligence , computer vision , projection (relational algebra) , computer science , tomosynthesis , pixel , iterative reconstruction , filter (signal processing) , preprocessor , maximum intensity projection , mammography , algorithm , medicine , radiology , cancer , angiography , breast cancer
Purpose: To generate a synthetic mammogram from a digital breast tomosynthesis (DBT) image which will be comparable to an actual mammogram. Methods: Prior to reconstruction of the DBT image, the obtained projections undergo preprocessing by modifying the distribution of pixel values in each projection. After reconstruction using filtered back‐projection with the processed projections, the forward projection and maximum intensity projection of the reconstructed image are taken using the ray‐driven scheme. The forward projection and maximum intensity projection are then put through a low‐pass and high‐pass filter respectively to obtain relevant information. The two filtered images are then added together to generate the synthetic mammogram. Results: Modifying the pixel value distribution in the projections prior to reconstruction improves contrast in the DBT image and therefore in the final synthetic image as well. This makes structures of interest such as microcalcification structures and mass lesions more readily identifiable. The downside to this approach is that noise is also amplified which can be partly mitigated by use of a low‐pass filter. In the synthetic image generated using the proposed method, structures of interest are just as visible as in the actual mammograms upon visual inspection. Combining the forward and maximum intensity projections after filtering gave better results that were visually more similar to the actual mammogram than using just the forward projection or maximum intensity projection alone. Conclusion: The results show that images generated using the proposed method are comparable to actual mammograms upon initial inspection, although further study of its diagnostic performance is needed.