ESTRO 2025 - Abstract Book

S641

Clinical - Breast

ESTRO 2025

4652

Digital Poster Predicting Tumour Response to Neoadjuvant Chemotherapy Using MRI Radiomic Features in Breast Cancer: a preliminary, bicentric analysis Berardino de Bari Department of Radio-oncology, Réseau Hospitalier Neuchâtelois, La Chaux-de-Fonds, Switzerland Purpose/Objective: Radiomics analyses offers a non-invasive tool to characterize breast cancer (BC) through advanced imaging techniques such as magnetic resonance imaging (MRI). By analysing the radiomic signature, this study aims to identify biomarkers predictive of tumour response to neoadjuvant chemotherapy (NAC) in patients with locally advanced BC. Material/Methods: This study included two cohorts: 40 patients from Hospital Universitario de Navarra (HUN, Spain) and 28 from Réseau Hospitalier Neuchâtelois (RHNe, Switzerland), all with histologically confirmed BC treated with NAC. Pre treatment breast MRI scans (SiemensÒ) were analysed, focusing on contrast-enhancing tumour regions. Regions of interest (ROIs) were contoured and refined by radiation oncologists and radiologists. For each case, ROI was defined as the contrast-enhancing tumour. A 1mm contraction of the ROI was performed to limit the analysis on the only tumour internal voxels and to account for potential heterogeneity in contouring. In case of multifocal breast cancers, patients with more than 3 tumour foci were excluded from the analyses. Using the z-score normalization, 163 radiomic features related to size, shape, and texture were extracted per IBSI standards. Correlations between these features and pathological complete response (pCR), observed in 31 cases (46%), were assessed using the Mann-Whitney test. Results: The most significant predictive biomarkers for pCR included Least Axis Length (p < 0.0058) and Cluster Shade (p < 0.0060), derived from the gray-level co-occurrence matrix. Other biomarkers with p <0.01 were excluded due to high Pearson correlation (>0.9) with these key predictors. Conclusion: This preliminary analysis highlights the potential of radiomics to support personalized NAC regimens, in order to optimise therapeutic outcomes while minimizing unnecessary toxicity. Our future work aims to expand the tumour volume analyses, incorporate advanced noise reduction techniques, and enhance dataset diversity. Integrating radiomics into BC management may revolutionize diagnosis, prognosis, and current treatment strategies, leading to improved patient outcomes.

Keywords: breast, radiomics, pathological complete response

References: [1] Li, H., Zhu, Y., Burnside, E. S., et al. (2018). Quantitative MRI radiomics in the prediction of molecular classifications of breast cancer subtypes in the TCGA/TCIA data set. NPJ Breast Cancer, 4, 17. [2 ]Braman, N. M., Etesami, M., Prasanna, P., et al. (2017). Intratumoral and peritumoral radiomics for the pretreatment prediction of pathological complete response to neoadjuvant chemotherapy based on breast DCE MRI. Breast Cancer Research, 19(1), 57.