RS-FISH was just released as open access article in @naturemethods
RS-FISH is a user-friendly software for accurate spot detection that is applicable to smFISH experiments, spatial transcriptomics, and spatial genomics. The approach enables fast spot detection in even very large volumetric datasets.
https://github.com/PreibischLab/RS-FISH
https://github.com/PreibischLab/RS-FISH-Spark
@preibischs @naturemethods Congratulations! But why the weird undersell in the size department?
> Owing to current limitations in Java arrays, the theoretical upper limit is 2^31 = 2,147,483,648 blocks, with each block maximally containing 2^31 = 2,147,483,648 pixels (for example 2048 × 2048 × 512 pixels). Given sufficient storage and compute resources, the limit for RS-FISH is thus 4,072 peta-pixels (4,072 petabytes at 8 bit, or 8,144 petabytes at 16 bit) taking into account the overhead, whereas every individual block locally processes only 2 gigapixels (2^31 = 2,147,483,648 pixels).
1. 2G^2 is 4exa, not 4peta
2. ImgLib2's `CellImg` manages 2^63 blocks, not 2^31, we had to change this for @davi 's [FAFB](https://linkinghub.elsevier.com/retrieve/pii/S0092-8674(18)30787-6) which had >2^31 blocks.
@herrsaalfeld @naturemethods @davi
* yes, 4,072 peta-pixels is ~4 exa-pixels - I just thought people can relate more to peta than exa.
* I did not know that 2^63 blocks are supported ... that increases that number significantly ... mmh
