search

DIAGNijmegen / picai_labels

Annotations for the PI-CAI Challenge: Public Training and Development Dataset
https://pi-cai.grand-challenge.org/
Other
50 stars 23 forks source link

ISUP 1 lesions #10

Closed edupooch closed 1 year ago

edupooch commented 1 year ago

Dear organizers,

I have been exploring the dataset for my research work, and I noticed there are 311 ISUP 1 lesions reported, but the expert-derived annotations for those weren't included in the dataset release. Were these samples not annotated or was the omission intentional to focus on the task of csPCa detection? Would it be possible to release the annotations for these samples as well?

Thank you for your hard work and dedication in putting together this dataset.

joeranbosma commented 1 year ago

Dear Eduardo,

Thanks for reaching out to us. The ISUP 1 lesions were only partially annotated, with many lesion masks missing (so while we did record their occurrence in the marksheet, we did not make voxel-wise lesion masks). This choice was made because they do not serve the primary objective of the PI-CAI challenge, but do take a lot of time to delineate.

I'm afraid the collection of these lesion masks would incur a lot of additional effort, which we rather spend differently.

All the best, Joeran

anindox8 commented 1 year ago

Indeed. To add upon @joeranbosma's remarks, some of these ISUP 1 lesions were really small (~1mm) and some were sampled via random systematic biopsies. So we couldn't reliably trace these histopathology findings back to their respective MRI scans, and create annotations with adequate certainty.