Abstract Funding Acknowledgements Horizon 2020 European Commission Project H2020-MSCA-ITN-2016 (764738), Grant from Fundacio La Marató de TV3 (040310) Background and aim Contemporary echocardiography provides complex data on cardiac function contained in both blood-pool and tissue deformation traces. Their current interpretation relies on clinical experience and selected peak or averaged velocity values, which might not capture the complexity throughout the cardiac cycle. The aim is to investigate if machine learning could recognize relevant patient profiles in arterial hypertension by integrating all echocardiographic data to better define potential pathophysiological changes in left ventricle (LV) remodeling. Methods An echocardiogram was performed in 100 patients with established arterial hypertension (> 3 years). Myocardial deformation traces of the LV and the left atrium (LA), assessed by 2D speckle tracking, the aortic outflow Doppler trace, the lateral and septal mitral annular Doppler velocity traces, and the mitral inflow Doppler trace were assessed as measures of cardiac function. An unsupervised machine learning algorithm (multiple kernel learning) was used to reduce the dimensionality of these data, and to position the patients based on the similarities of echocardiographic data. The main patterns of variability present in the data were interpreted through non-linear regression analysis. Classic echocardiographic parameters, measured by a clinician, were then compared between the intermediate and extreme patient profiles across the variability spectrum. Results Figure 1 shows differences in velocity and deformation traces between the three representative patient profiles. While at one end of the spectrum, all echocardiographic traces were normal (red and green), the data of the other extreme patient profile (blue) describes a characteristic and consistent LV pressure overload remodeling pattern, with slightly reduced and delayed aortic outflow velocities, fused E and A waves with the ratio < 1, lower e’ mitral annulus velocities, decreased basal septal strain with post-systolic motion, prolonged relaxation in early diastole as seen by the deformation traces, and a change in atrial deformation dynamic with augmentation of LA contractile strain. The clinical measurements concurred with the remodeling profile describing smaller end-diastolic LV diameter and end-systolic and end-diastolic LV volumes, a reduced E/A ratio and e’ medial annular velocity, reduced TAPSE and increased LA contractile strain. Conclusion Machine learning based assessment of complex echocardiographic data has the potential to recognize an integrated and comprehensive patient profile related to LV remodeling within the hypertensive cohort without relying on classical clinical measurements and parameters, but by learning from subtle differences globally present in velocity and deformation echocardiographic data. Abstract 421 Figure 1