We have noticed errors in mask outputs for certain datasets such as high-density tissue or 5um tissue modes where the number of mask objects can exceed the max for unsigned 16-byte integers of 65535. For example, masks may start to skip integers for mask ID objects which can cause errors in downstream analysis.
Would be preferred if the default mask dtype on export would be unsigned 32 byte integers so that these masks with many objects can be processed consistently.
matt-sd-watson
commented
3 weeks ago
We have noticed errors in mask outputs for certain datasets such as high-density tissue or 5um tissue modes where the number of mask objects can exceed the max for unsigned 16-byte integers of 65535. For example, masks may start to skip integers for mask ID objects which can cause errors in downstream analysis.
Would be preferred if the default mask dtype on export would be unsigned 32 byte integers so that these masks with many objects can be processed consistently.