Association of echocardiographic radiomics-based features with cardiotoxicity effect in breast cancer patients from the CARDIOCARE project

Title: Association of echocardiographic radiomics-based features with cardiotoxicity effect in breast cancer patients from the CARDIOCARE project Authors: G. Manikis, G. Kalliatakis, K. Marias, V. Bouratzis, L. Lakkas, A. Naka, G. Karanasiou, D. Tsekoura, E. Kampouroglou, K. Keramida, A. Papakonstantinou, A. Constantinidou, K. Mazzocco, C. Cipolla, D. Fotiadis Affiliations: 1. Foundation for Research and Technology - Hellas, Heraklion, Greece 2. University hospital of Ioannina, Ioannina, Greece 3. University of Ioannina, Ioannina, Greece 4. Aretaieio Hospital, Athens, Greece 5. Attikon University Hospital, Athens, Greece 6. Karolinska Institutet, Stockholm, Sweden 7. University of Cyprus, Nicosia, Cyprus 8. University of Milan, Milan, Italy 9. European Institute of Oncology, Milan, Italy On behalf of CARDIOCARE EU project Manuscript: The manuscript explores the association between echocardiographic radiomics-based features and cardiotoxicity in breast cancer patients from the CARDIOCARE project. Improvements in breast cancer therapy have led to increased survival, but also more cardiotoxic effects like cardiomyopathy and heart failure. The study aims to leverage radiomics analysis for non-invasive echocardiography biomarkers associated with cardiotoxicity. In a prospective cohort of 28 patients (above 60 years old, 10 developed cardiotoxicity) undergoing antineoplasmatic therapy, over 1500 imaging features were extracted from echocardiographic images (apical 4and 2-chamber views) at baseline and after 3 months using a pre-trained neural network for segmentation. Cardiotoxicity was assessed by combining imaging (LVEF, GLS) and blood markers (Troponin I, Nt-proBnp). Univariate analysis showed 40 textural radiomics features (mostly from the myocardium, especially in the 4-chamber end-diastolic view and 2-chamber endsystolic view) were significantly associated with cardiotoxicity, while histogram and shape features were not. The study concludes that myocardial heterogeneity quantified by radiomics could have diagnostic value for early detection of cardiac toxicity.