Back to the list
Congress: ECR25
Poster Number: C-11257
Type: Poster: EPOS Radiologist (scientific)
Authorblock: M. Swoboda, J. Deeg, D. Egle, V. Ladenhauf, M. Galijašević, S. Haushammer, B. Amort, M. Pamminger, L. Gruber; Innsbruck/AT
Disclosures:
Michael Swoboda: Nothing to disclose
Johannes Deeg: Nothing to disclose
Daniel Egle: Nothing to disclose
Valentin Ladenhauf: Nothing to disclose
Malik Galijašević: Nothing to disclose
Silke Haushammer: Nothing to disclose
Birgit Amort: Nothing to disclose
Mathias Pamminger: Nothing to disclose
Leonhard Gruber: Nothing to disclose
Keywords: Artificial Intelligence, Breast, Oncology, Elastography, Ultrasound, Diagnostic procedure, Cancer
Results

A total of 363 benign, 18 intermediate and 40 malignant lesions were analysed. The majority of tumors were benign (86.2%, n = 363), with a smaller proportion exhibiting intermediate (4.3%, n = 18) or malignant differentiation (9.5%, n = 40). The most common benign tumors included fibroadenomas (90.9%, n = 319), fibrocystic mastopathy (7.1%, n = 25), and adenosis (2.9%, n = 10). Algorithm based quantitative ranking showed that the most predictive features indicating malignancy were hyperechoic rim (gain ratio merit 0.135 ± 0.004), irregular border (0.057 ± 0.002), perilesional stiffening (0.054 ± 0.002), pectoral contact (0.051 ± 0.003), irregular shape (0.029 ± 0.001) and irregular vasculature (0.027 ± 0.002). Notably, patient age, lesion size (in all dimensions), volume, height-to-width ratio, and volume doubling time did not demonstrate significant value for classification purposes.

GALLERY