Back to the list
Congress: ECR25
Poster Number: C-13257
Type: Poster: EPOS Radiologist (scientific)
Authorblock: C. Piccolo, M. Sarli, M. Pileri, M. Tommasiello, A. Rofena, V. Guarrasi, P. Soda, B. Beomonte Zobel; Rome/IT
Disclosures:
Claudia Piccolo: Nothing to disclose
Marina Sarli: Nothing to disclose
Matteo Pileri: Nothing to disclose
Manuela Tommasiello: Nothing to disclose
Aurora Rofena: Nothing to disclose
Valerio Guarrasi: Nothing to disclose
Paolo Soda: Nothing to disclose
Bruno Beomonte Zobel: Nothing to disclose
Keywords: Breast, Mammography, Computer Applications-Detection, diagnosis, Cancer
Results

Univariate Analysis: Radiomic Features and Prognostic FactorsThis study highlights the potential of radiomics in breast cancer care by analyzing features from both tumor lesions and their contours. Radiomic features were shown to provide clinically relevant information for diagnosis, risk stratification, and treatment planning. Importantly, the findings emphasize the need for an integrated, multi-dimensional radiomic approach to maximize its utility in personalized treatment strategies and improve patient outcomes.

Lesion Contour Features: Strong Correlations with Prognostic FactorsRadiomic features extracted from lesion contours demonstrated stronger correlations with prognostic factors compared to features from lesions alone. For instance, Uniformity (ER, r = 0.4719, p = 0.0009) and Gray Level Non-uniformity Normalized (PgR, r = 0.4792, p = 0.0008) had Pearson coefficients exceeding 0.4, suggesting their value in predicting tumor behavior. These findings underscore the clinical significance of contour analysis in offering additional biological insights and guiding management strategies (Table 1).

Multivariate Analysis: Combining Feature Classes for Greater InsightMultivariate analysis revealed that combining features from both CC and MLO images resulted in stronger statistical correlations with prognostic factors compared to using either imaging perspective alone. As shown in Figures 1–3, feature classes such as GLCM and NGTDM demonstrated significant associations with factors like ER and PgR. This highlights the importance of integrating data from multiple imaging perspectives to better understand tumor behavior. Key results, including R² values and p-values, are summarized in Table 2.

Lesion Contour Features in Multivariate AnalysisLesion contour features also showed significant correlations with prognostic factors in multivariate analysis, even when using CC or MLO images separately. Notably, all feature classes (except Shape Features) exhibited significant associations with HER2, with p-values often below 0.0001. These results, detailed in Tables 3–5, highlight the potential of radiomics to complement traditional diagnostics, particularly in cases where conventional methods face limitations.
GALLERY