KiTS19 Changelog

April 15 -- Amended Training Data Release

Improved visualization
- visualize.py now allows for a --plane argument (-p) where you can specify axial, coronal, or sagittal.
- The pixels will be interpolated to an aspect ratio consistent with the real world in all cases.
Cropped cases 22, 61, and 117
- These had a second series at a different contrast phase appended to the end of their z-dimensions (#6).
Parapelvic cysts and hydronephrosis were corrected
- Case 155 had a cyst that was originally mistaken for hydro (#10).
- Case 166 had hydro that was originally mistaken for a cyst (#11).
Fixed holes in tumor and kidney annotations
- Thresholds were adjusted to avoid the exclusion of hypodense tissue in kidney and tumor interiors (#12).
Four cases were replaced entirely
- The tumors in cases 60 and 174 were complex cysts. The decision was made to exclude these from our cohort to avoid ambiguity with other cysts labeled as kidney.
- The tumor in case 15 was actually a blown out lower pole.
- The tumor in case 102 was actually an adrenal cortical carcinoma abutting the kidney.
The pelvic kidney in case 40 was annotated
- This was missed in the original release
The hilum issues in case 42 were corrected
- On both the left and right side, the upper pole hilums were not handled correctly.
The affine matrices were corrected
- In the initial release, the slice thickness was incorrect for many thin-cut cases (#8).
- In the initial release, the affine matrices were incorrectly flipped along the horizontal axis (#9).

210 Cases in original spacing were added, along with some starter code to visualize the imaging and segmentation labels in the axial plane.