Medical data continues to grow incrementally and it is useful to have fixed versions to look back on for benchmarking and communicating our results.
If you want to make a new version of the dataset, please follow the rules for versioning detailed here (link to be added). Please also run your versioning by both Arjun Desai (arjundd at stanford dot edu) and Akshay Chaudhari (akshaysc at stanford dot edu).
The most up-to-date version of the qDESS dataset is v1.0.0.
Version history is described below. If new versions are made, add details appropriately
Date Released: August 23, 2021
This is the first stable release of the qDESS dataset.
Changelist:
- Bumping
v0.1.0 -> v1.0.0
to indicate stable release. - No changes were made; this version is synonymous with
v0.1.0
.
Date Released: March 8, 2021
Splits are the same as v0.0.2. Image data can be found at:
/bmrNAS/people/arjun/data/qdess_knee_2020/image_files_v0.1
Changelist:
-
Segmentation masks in the test set were corrected by two additional annotators. The following segmentations were changed:
MTR_005
,MTR_030
,MTR_066
,MTR_134
,MTR_219
,MTR_248
. -
Additional bounding box annotations
Data Distribution:
# Scans % Male % Female Age (mean +/- std) # BBoxes
----- --------- -------- ---------- -------------------- ----------
train 86 53.5 46.5 43.9 +/- 18.2 242
val 33 54.5 45.5 44.5 +/- 18.1 104
test 36 72.2 27.8 42.1 +/- 15.8 130
Date Released: November 24, 2020
Splits are the same as v0.0.1. Additional bounding box annotations were added.
# Scans % Male % Female Age (mean +/- std) # BBoxes
----- --------- -------- ---------- -------------------- ----------
train 86 53.5 46.5 43.9 +/- 18.2 239
val 33 54.5 45.5 44.5 +/- 18.1 102
test 36 72.2 27.8 42.1 +/- 15.8 130
Date Released: October 16, 2020
This is the first version of the qDESS dataset.
Examples in the test split have received arthroscopic surgery (to establish gold standard injury reports) and were evaluated by 2 msk rads. Training and validation splits were pseudo randomly determined from the remaining scans while approximiately balancing age and sex. Details are shown below:
# Scans % Male % Female Age (mean +/- std) # BBoxes
----- --------- -------- ---------- -------------------- ----------
train 86 53.5 46.5 43.9 +/- 18.2 207
val 33 54.5 45.5 44.5 +/- 18.1 88
test 36 72.2 27.8 42.1 +/- 15.8 23