Published March 2, 2021 | Version 1.0.0
Other Open

Curated annotations for SpineWeb Dataset 16

Authors/Creators

  • 1. Friedrich-Alexander-Universtität Erlangen-Nürnberg, Erlangen, Germany

Description

This repository provides curated annotation data for the SpineWeb Dataset 16 [1,2].

The original annotation data has label noise:

  • Upward/Downward shift of the vertebral body annotations.
  • Flipping of corner points and top/end plates. 
  • Swap of vertebral annotation, e.g. L5 at position of T1.
  • Overlapping vertebrae. 
  • Corner point duplicates.
  • Anatomically implausible corner point positions.

First, we performed a couple of automatic sanity checks:

  1. Check that the midpoints of each consecutive vertebra is below the previous one to ensure anatomically correct spine configurations.
  2. Check that ordering of the points is correct for each vertebra.
  3. Check for overlay of multiple vertebrae.
  4. Check for the total number of vertebrae.

Following identification of the most problematic cases, we performed manual corrections where needed. The corrections done and problem descriptions for each case are available as a spreadsheet.

The corrected/curated annotations are stored in '.xml' files (one for each image with matching file name and xml meta information) and are sorted w.r.t. the vertebra index. For each vertebra (e.g. L1), the 4 corner points are annotated as a list [x1,y1, x2,y2, x3,y3, x4,y4]. The ordering is top-left, top-right, bottom-right, bottom-left. 

 

[1] Wu, H., Bailey, Chris., Rasoulinejad, Parham., and Li, S., 2017.Automatic landmark estimation for adolescent idiopathic scoliosis assessment using boostnet. Medical Image Computing and Computer Assisted Intervention:127-135.

[2] http://spineweb.digitalimaginggroup.ca/Index.php?n=Main.Datasets

 

 

Files

annotations_test.zip

Files (729.5 kB)

Name Size Download all
md5:d2781392330c4142d9ad35550e1eae10
106.0 kB Preview Download
md5:03c7fa5524063ee2d8862bf8b45cc728
396.7 kB Preview Download
md5:93bbb2be63c197fbb44fcdca601f9b21
164.7 kB Download
md5:e36e36f35ca4fc014ccaa49f03d6fcbf
19.4 kB Preview Download
md5:810b4d6f5816abf9c1cedac627644fc2
42.6 kB Preview Download

Additional details

Related works

Is derived from
Conference paper: 10.1007/978-3-319-66182-7_15 (DOI)
Is part of
Conference paper: 10.1007/978-3-658-33198-6_85 (DOI)