Generating Synthetic Mammograms From Reconstructed Tomosynthesis Volumes

Area of Science:

• Radiology
• Medical Imaging

Background:

• Digital breast tomosynthesis (DBT) offers 3-D imaging but may increase lesion search time and complicate comparisons.
• Synthesizing 2-D mammograms from DBT could streamline screening and guide abnormality detection.

Purpose of the Study:

• To develop and evaluate a method for synthesizing 2-D mammograms from DBT volumes, optimizing visualization of masses and architectural distortions.
• To assess the diagnostic performance of synthesized mammograms compared to conventional mammograms for mass lesion detection.

Main Methods:

• A computer-aided detection system identified relevant points in DBT volumes.
• A surface was fitted through these points, and mammograms were rendered at the intersection with the DBT volume.
• A pilot observer study compared mass detection in synthesized mammograms versus conventional mammograms.

Main Results:

• Diagnostic accuracy for mass detection was slightly higher in synthesized mammograms (Az=0.85) compared to conventional mammograms (Az=0.81).
• The difference in diagnostic accuracy was not statistically significant.
• Synthesized mammograms were found to be an acceptable alternative for detecting mass lesions.

Conclusions:

• Synthesized 2-D mammograms from DBT show potential as a viable alternative to conventional mammograms for mass lesion detection.
• This technique may help overcome the challenges of lesion search and comparison in DBT screening.
• Further validation is suggested for clinical implementation.