Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

Voting strategy for artifact reduction in digital breast tomosynthesis.

Tao Wu1, Richard H Moore, Daniel B Kopans

  • 1Massachusetts General Hospital, Boston, Massachusetts 02114, USA. twu@hologic.com

Medical Physics
|August 11, 2006
PubMed
Summary

This study explores new techniques to improve image clarity in 3D breast scans by identifying and removing visual noise caused by dense objects like calcifications. Researchers developed a voting strategy that selectively ignores specific data angles that would otherwise create distracting streaks or shadows in the final image. By comparing four different mathematical approaches, the team found that combining specific segmentation techniques with classification methods provides the most effective way to clean up images. These findings help radiologists see breast tissue more clearly by minimizing errors caused by high-density features.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Correcting Decades of Misinformation About Breast Cancer Screening: An Open Letter to Women and Those Who Advise Them About Screening for Breast Cancer.

Journal of breast imaging·2025
Same author

More than a Half Century of Misinformation About Breast Cancer Screening.

Radiologic clinics of North America·2024
Same author

The New Proposed U.S. Preventive Services Task Force Recommendation on Breast Cancer Screening for Women in Their 40s.

Annals of internal medicine·2024
Same author

Antecedents: A Half-Century of Imaging the Breast.

Journal of breast imaging·2024
Same author

Breast cancer outcomes based on method of detection in community-based breast cancer registry.

Breast cancer research and treatment·2023
Same author

Including the method of detection for breast cancer in the Surveillance, Epidemiology, and End Results database is long overdue.

Journal of medical screening·2023

Area of Science:

  • Medical imaging physics within digital breast tomosynthesis
  • Radiological diagnostic instrumentation and signal processing

Background:

Limited angular range and sparse projection counts often lead to image degradation in clinical breast imaging. No prior work had resolved how to systematically mitigate these specific reconstruction errors. It was already known that dense anatomical structures frequently introduce distracting streaks into three-dimensional volumes. This uncertainty drove the need for more robust computational filtering techniques. Prior research has shown that high-attenuation objects create significant challenges for standard back-projection algorithms. That gap motivated the development of novel strategies to handle these problematic data points. Researchers have long sought ways to improve diagnostic accuracy by refining how raw data is processed. This study addresses these persistent challenges by proposing a new framework for selective data rejection.

Purpose Of The Study:

The study aims to develop and evaluate several artifact reduction methods for digital breast tomosynthesis. Researchers sought to address the persistent problem of reconstruction artifacts caused by high-attenuation features within breast tissue. The team investigated how a voting strategy could identify and reject projections that introduce visual noise. This work was motivated by the limitations inherent in systems with narrow angular ranges and sparse projection counts. The authors intended to compare four specific approaches to determine their relative effectiveness in improving image clarity. They specifically examined how different algorithms handle both large and small calcifications. By testing these techniques, the researchers hoped to provide a solution for minimizing distracting streaks in clinical volumes. This investigation establishes a framework for more accurate and reliable breast imaging reconstructions.

Keywords:
streak reductioncalcification detectionmedical imaging physicstomographic reconstruction

Frequently Asked Questions

The researchers propose a voting strategy that identifies and rejects projections contributing to voxel artifacts. This mechanism selectively ignores data from specific angles when high-attenuation features are detected, preventing the propagation of streaks into the final reconstructed volume.

The team compared four distinct approaches: projection segmentation, maximum contribution deduction, one-step classification, and iterative classification. Each method uses different criteria, such as segmenting high-attenuation features or comparing contribution values, to determine which projections should be excluded during reconstruction.

Segmentation is necessary for handling metal and large calcifications because these features can be reliably detected and isolated from raw projections. This approach allows the system to specifically target and remove data that would otherwise cause significant image distortion.

Related Experiment Videos

Main Methods:

The review approach evaluated four distinct computational strategies for filtering projection data during volume generation. Researchers implemented a voting framework that dynamically assesses the impact of individual projections on each voxel. The team utilized projection segmentation to isolate high-attenuation objects directly from raw data streams. They also tested three classification-based algorithms that compare contribution values across different angular views. Each method was systematically compared to determine its efficacy in suppressing streak artifacts. The study design focused on identifying projections that introduce noise rather than signal. Quantitative comparisons were performed to assess the trade-off between artifact removal and potential image degradation. This analysis provided a rigorous assessment of how different mathematical models handle dense anatomical features.

Main Results:

The iterative classification method provides the most effective artifact reduction among the tested approaches. However, this specific technique also generates numerous false positive classifications that negatively impact overall image quality. The maximum contribution deduction and one-step classification methods both successfully reduce artifacts from small calcifications. One-step classification demonstrates slightly superior performance compared to the maximum contribution deduction method. Projection segmentation effectively targets and removes artifacts caused by metal and large calcifications. Combining one-step classification with projection segmentation successfully eliminates artifacts from both large and small calcifications. These findings demonstrate that hybrid strategies outperform individual methods in diverse clinical scenarios. The results highlight the necessity of balancing noise suppression with the preservation of diagnostic information.

Conclusions:

The iterative classification approach yields the most significant reduction in visual noise but risks introducing false positives. One-step classification performs better than maximum contribution deduction for small calcifications. Combining projection segmentation with one-step classification effectively handles both large and small high-attenuation features. These results suggest that selective data rejection improves overall image fidelity in tomosynthesis. The authors propose that segmentation is highly reliable for identifying metal or large calcifications. The study highlights trade-offs between artifact suppression and potential image degradation. These findings imply that hybrid methods offer the most balanced performance for clinical applications. Future implementations could benefit from refining the classification thresholds to minimize false positive rates.

Projection data serves as the input for the voting strategy, where individual contributions are analyzed to detect potential artifacts. The classification components then act as the decision-making layer, determining whether to include or reject specific projections based on their contribution values.

The researchers measured artifact reduction performance across different calcification sizes. They observed that while iterative classification provides the best suppression, it also generates more false positives compared to the one-step classification method, which performs better for small calcifications.

The authors suggest that combining one-step classification with projection segmentation provides the most comprehensive solution. They propose this hybrid approach effectively removes artifacts from both large and small calcifications, offering a more robust outcome than any single method alone.