Ultrasonic fat fraction quantification using in vivo adaptive sound speed estimation

The non-invasive quantification of human tissue fat fraction using easily scalable and accessible imaging technologies is crucial for the diagnosis of many diseases including liver steatosis. Here, we propose a non-invasive quantification of fat content using a highly accessible ultrasonic imaging technology. Ultrasonic echoes backscattered from human liver tissues are recombined to synthetize echoes of a virtual point-like reflector within the organs. This virtual point-like reflector is an ultrasonic analogue of artificial stars generated by laser beams in the field of astronomy, which are used to estimate the aberrations induced in the propagation medium. Here, the ultrasonic echoes from the point-like reflector provide an estimate of the Green's function relating the ultrasonic array and the reflector location and consequently represent a measurement of the aberrations induced along the ultrasonic beam travel path. Maximizing the spatial coherence of echoes backscattered from this targeted region provides an estimate of the acoustic sound speed while iteratively making the reflector more echogenic. The acoustic sound speed is dependent of the organ fat content, and we derive and cross-validate a theoretical equation relating acoustic sound speed and fat content both in phantom experiments and humans. An ultrasound-based fat fraction was found to be highly correlated with the oil paraffin concentration (R ²  =  0.985) in phantoms and well correlated with the gold standard magnetic resonance imaging proton density fat fraction measurements (R ²  =  0.73) in patients.