A method and device for detection of bone in meat identifies fragments larger than about 1 mm using spectral optical imaging and ultrasound. Spectral imaging can detect foreign material proximate to the surface and ultrasound can detect material within the sample. The sample is irradiated by light and reflected light or Raman scattered light measured. The sample is similarly irradiated by ultrasound and reflected or transmitted sound waves give a set of amplitude data points, which include temporal delay. These data points are then processed by statistical methods to derive a set of vectors in n-dimensional space, which are compared to a calibrated data set of derived vectors which have distinct identifying loci for each type of surface, are indicative of the presence or absence of defects.