Modeling lesion malignancy in in‑silico breast imaging

Determine a validated computational method to model the malignancy of breast mass lesions within the VICTRE in silico imaging pipeline and lesion growth framework, so that malignancy can be represented explicitly (for example via spiculation or other radiological features) and used consistently in longitudinal virtual trials of digital mammography and tomosynthesis.

Background

The authors previously developed a fully in silico clinical trial that replicated a real breast cancer detection trial and later expanded it to a longitudinal design with growing lesions. Although they introduced a spiculation model with tunable parameters (length, thickness, and density) and performed a reader study, they acknowledge that explicitly modeling malignancy itself remains unresolved.

Spiculation is recognized as a biomarker associated with malignancy, but a comprehensive, validated approach to encode and quantify malignancy in simulated lesions—beyond spiculation alone—has not yet been established. Addressing this would enable more clinically faithful virtual trials that include malignancy classification tasks.