Traditionally diagnostic ultrasound Imaging is performed using simplified physical-model for wave propagation and with assumptions for physical tissue properties, like the tissue speed of sound. While this may be sufficient to produce diagnostic images for qualitative human evaluation, the resolution and contrast of such methods can be improved by applying machine learning to certain processing steps in the imaging pipeline or as an end-to-end solution for simultaneously and jointly predicting the quantitative tissue properties while improving the resulting image quality. The former has further diagnostic value as an imaging biomarker – hence forms a novel approach for a wide range of diseases, including (but not limited to) Non-Alcoholic Fatty Liver Disease (NAFLD), breast cancer detection, and sarcopenia screening. In this project we will develop and train ML models for better ultrasound image beamforming and for local tissue properties reconstruction, such as for the quantification of backscattering, speed-of-sound, and attenuation)