Breast, Computer applications, Artificial Intelligence, Mammography, RIS, Image manipulation / Reconstruction, Screening, Statistics, Biopsy, Cancer, Image verification, Pathology
F. Leone1, A. Presazzi2, M. Cellina1, M. A. Orsi1, G. Oliva1; 1Milan/IT, 2PAVIA, IT/IT
Methods and materials
We reviewed a retrospective cohort of 1530 women patients with clinical Mammograms between April 2018 and September 2018 in an Italian Hospital.
All the mammograms have been performed by the same device.
We included the patients with breast cancer mass histopathology proved (we excluded microcalcificantion lesions) and images in full compliance with standard quality.
We evaluated two different groups of images: first group made of mammograms with breast cancer and the second group made of controlateral mammograms without brast cancer and neither suspicius lesions.
We obtained Radiomic's data drawing ROI of the all mammogram,
avoiding pectoral muscles,
macrocalcifications and the nipple.