(no subject)

[timelets]

But it was the idea of a pattern of differentially attenuated ultrasound that occurred to Karl Theo and Friedrich Dussik, Austrian brothers, one a neurologist and the other a physicist, in the late 1930s. In other words, differences in the amount of energy transmitted through an organ could be used to create a pattern that would represent the form of that organ. This led the Dussiks to make the first claim that ultrasound could be used diagnostically.

In 1937 he [Karl Theo Dussik] obtained the first pictures of the patterned attenuation of an ultrasound beam passing through the skull at a series of positions. The intention was to discover abnormalities in the shape of the ventricles in the brain without using x-rays and dyes. The energy of the transmitted beam could be registered on a photographic plate, so that as the transmitter was moved across the skull a pattern of dark and light patches was built up on the plate.
...

Karl Dussik’s work came to the attention of the American physicist, Richard Bolt, who was director of the Acoustics Laboratory at MIT from 1946 to 1957.

The medical acoustics project was just one of a number of projects, although it is the one with which Bolt was involved. In April 1949 Bolt, his colleague Leo Beranek, and their collaborator at the Massachusetts General Hospital, the brain surgeon H. Thomas Ballantine, wrote to J. R. Killian, the president of MIT, seeking funds for their proposed research.

-- Edward Yoxen. Seeing with Sound: A Study of the Development of Medical Images

The same principle was later used in MRI machines. In general, the idea in combination with calibration seems to be "deeply learnable," although you can completely skip the visualization part. Essentially, we induce and map differences in one domain into a perceivable codomain in order to extract objects/processes and map them further to diagnostics info.

Related: https://a16z.com/2017/12/14/bio-team-bio-fund-ii/