I read up on the Wikipedia page on this type of imaging and also viewed the Crowd Supply site video and read everything there. While the technology is fascinating and has amazing potential, it will probably find niche applications in screening and other situations in which low-cost and safety are important, but will likely have high false positive rate... Basically you want low specificity, but also sensitive enough not to miss pathology when it is there, to push people to then move on to further testing with other more expensive but proven medical imaging technologies that use more penetrating forms of radiation and can provide higher resolution data.
Part of the problem I see is that the conductive path is bent and distorted by tissue. Sure in a water tank it would be more linear because of the homogenous medium, but entering the body you have various composition of muscle, fat, ligaments, nerves, etc. So for direct visualization of enough resolution that you can use to spacially measure anatomy I don't think it can be applied the same way a CT or MRI can. Even ultra-sound can suffer from this issue. Using it to build a "controller" where you can tell the computer to memorize different patterns to mean different hand-gestures is a different story. You get a funny looking blob hear, it means one thing.... a funny looking blob there, it means another. As long as you get consistency you can pattern match and interpret the appropriate gesture. But the minute you move it to another person, or even the same user and position the sensor array differently, the patterns are all going to be different and you have to "recalibrate".
I would be curious to see what results in the test tank if you change around the distribution or use combinations of carrier medium (oil, water, glycerin, Jello, etc...) and immerse substances of either similar electrical impedance or different conductivities, or adding electrolytes to the water? As far as comparing to a CT scan, my understanding is that the imaging sensor works by producing current at one electrode while listening with the others, and steps through each one likewise, continuously circling around, thus building up a map of the space inside of the ring of electrodes. A CT machine does the same but because it has a radiation source at one point and imaging plate at another, the entire machine has to circle around the object. With electrodes, each one can be turned into an "emitter" and "sensor" so you don't need to move the entire structure around, you just have software alternately choose and circle around in a loop-like fashion like this:
Assuming 10 electrode circle:
for e=1 to 10 do: // circle through each electrode
{
for s=1 to 10 do: // listen through each electrode
if S=E do nothing, // except for when it is same electrode, skip
else:
{
pulse from electrode[e]; // turn on a pulse at electrode "e"
sense at electrode[ s]; // read at electrode "s"
}
}
Or do you read all the other sensing electrodes simultaneously?
Now you read the values you get doing each "pairwise" selection of electrodes[e,s] and work backwards to geometrically color in a grid by filling in the value along a line between the same electrodes in the same spacial orientation you have. You basically integrate the values on top of that grid/map and then normalize it (or color map) to show the differences with more contrast.
What I'd like to see is the same thing but perhaps instead of just electrodes (for impedence), place piezo-electrics in the same array and see if you can use vibration/sound to image something. Perhaps you can get unique data out of combining both or doing other types of arrays employing the use of multiple physical attributes in measurements.
Fascinating, can't wait to learn more and see it in action.