Read one region of a giant slide
A whole-slide microscopy image (NDPI / SVS / TIFF) is a gigapixel pile of JPEG tiles, mostly empty background. To crop one tissue region, a native reader must download the whole file and decode it. With VirtualiZarr and Icechunk, a tiny manifest maps every tile to a (offset, length) in the slide. The reader range-GETs only the region's tiles, one coalesced read per tile-row, and the image fills in as it arrives.
Whole-slide reader bill
$0
< 1 s
0 B downloaded
Virtual chunks bill
$0
< 1 s
0 B downloaded
☁️ Object storage — whole-slide image
Native whole-slide reader
Your laptop
download + decode whole file
cropped region
held on laptop
0 B
Icechunk virtual dataset
Icechunk repo
manifest
(offset, length) per tile
Your laptop
region, row by row
held on laptop
0 B