Hybrid simulation of breast CT for assessing microcalcification detectability.

Area of Science:

• Medical Imaging
• Radiology
• Computational Imaging

Background:

• Virtual imaging trials (VITs) are crucial for efficient regulatory evaluation of novel imaging technologies.
• Breast computed tomography (CT) requires robust methods for assessing microcalcification detection.

Purpose of the Study:

• To develop and validate a hybrid VIT methodology for breast CT.
• To investigate the detectability of microcalcifications using this hybrid VIT approach.

Main Methods:

• Generated simulated microcalcification projection images using ray tracing.
• Integrated simulated images with real patient breast CT data.
• Employed human observers (HOs) and deep learning model observers (DLMOs) for detection analysis.
• Utilized receiver operating characteristic (ROC) curve analysis.

Main Results:

• DLMOs achieved high AUC values (0.80-0.99) for detecting microcalcifications of varying sizes.
• Both HOs and DLMOs demonstrated strong performance for larger microcalcifications (≥0.21 mm).
• Microcalcification conspicuity was higher in adipose tissue and anterior breast locations.

Conclusions:

• The developed hybrid VIT methodology is promising for assessing breast CT imaging systems.
• This approach enables comprehensive evaluation across diverse parameters.
• Hybrid VITs can significantly enhance the evaluation of imaging technologies.