Fig 1. Puzzle Imaging.

There are many properties, such as genetic information, that are easier to determine when the original spatial information about the sample is lost. However, it may be possible to still image these properties using relative spatial information. (A) As an example, let us say that each piece of genetic information is attached to a puzzle piece. While the puzzle pieces don’t provide absolute spatial information, they provide relative spatial information: we know that nearby pieces should have similar colors, so we can use color similarity to determine how close puzzle pieces should be to one another. (B) We can make a similarity matrix of the puzzle pieces, which states how similar the puzzle pieces’ colors are to each other, and thus how close the pieces should be to one another. (C) Through dimensionality reduction techniques, this similarity matrix can be used to map each puzzle piece to its correct relative location.