A cross-cutting approach for tracking architectural distortion locii on digital breast tomosynthesis slices

摘要: Abstract Background and objective: Full-field digital mammography (FFDM) is the predominant breast cancer screening exam used. However, with emergence of tomosynthesis (DBT) radiologists could improve early recognition signs. In this scenario, detection architectural distortion (AD) still a challenging task. ADs are very subtle contraction parenchyma that represent earliest manifestation cancer, assessing at present 50% missed cases. Methods: This paper proposes new paradigm to detect AD in DBT exams by cross-cutting approach exploiting 3-dimensionality imaging modality. After locating candidates each slice, suspicious spots tracked cross-slice direction then characterized terms neighboring texture. approach, which mimics radiologist's scrolling down over zoomed slices, we reduce amount uninformative signs collected preserving large variability appearance. Results: Using 37 sets slices containing least one locus indicated radiologist, proposed methodology reaches an AUC 0.84, only false negative sensitivity 0.9. Conclusions: The results show algorithm can be promising tool for automatic locii. Future work will address extension dataset as well improvement performance toward application clinical practice.