Reference Datasets

All of the datasets available through this website have previously been reported as part of peer-reviewed publications (see references below). They can be obtained via the Downloads page of the website. Currently, the set of available reference data consists of:

  1. Thoracic 4DCT images acquired as part of the radiotherapy planning process for the treatment of thoracic malignancies at the University of Texas M. D. Anderson Cancer Center in Houston, TX, and
  2. Inspiratory and exipratory breath-hold CT image pairs acquired from the National Heart Lung Blood Institute COPDgene study archive.

Below is a summary of the image and feature characteristics for the currently available reference datasets. We will update the library as more cases are added, and will inform all of our registered users as they become available.

Table 1. DIR-Lab Reference Datasets
  • 4D CT
  • COPDgene
Table 1a. Reference Datasets - 4DCT. Each case is identified according to the given label. The image dimensions are given in voxel units, and the voxel dimensions are given in millimeters. The "# Features" column designates the total quantity of unique landmark features identified for each case. For the 4DCT images, the full point set is identified between the maximum inhalation and exhalation component phase images. Additionally, a subset of 75 features has been propagated onto each of the expiratory phase images (i.e., T00, T10, T20, T30, T40, and T50). The "Displacement" column shows the mean (and standard deviation) displacement of the complete primary feature set. The entries in the "# Repeats" column are formatted as (Nm / Nobs), where Nm is the number of repeat registration measurements performed by each of Nobs independent observers. The "Observers" column shows the combined mean (and standard deviation) repeat registration error for the set of Nobs data sets. Please see the references cited below for more information.
Label Image Dims Voxels (mm) # Features Displacement (mm) # Repeats Observers (mm)
4DCT1 256 x 256 x 94 0.97 x 0.97 x 2.5 1280 4.01 (2.91) 200/3 0.85 (1.24)
4DCT2 256 x 256 x 112 1.16 x 1.16 x 2.5 1487 4.65 (4.09) 200/3 0.70 (0.99)
4DCT3 256 x 256 x 104 1.15 x 1.15 x 2.5 1561 6.73 (4.21) 200/3 0.77 (1.01)
4DCT4 256 x 256 x 99 1.13 x 1.13 x 2.5 1166 9.42 (4.81) 200/3 1.13 (1.27)
4DCT5 256 x 256 x 106 1.10 x 1.10 x 2.5 1268 7.10 (5.14) 200/3 0.92 (1.16)
4DCT6 512 x 512 x 128 0.97 x 0.97 x 2.5 419 11.10 (6.98) 150/3 0.97 (1.38)
4DCT7 512 x 512 x 136 0.97 x 0.97 x 2.5 398 11.59 (7.87) 150/3 0.81 (1.32)
4DCT8 512 x 512 x 128 0.97 x 0.97 x 2.5 476 15.16 (9.11) 150/3 1.03 (2.19)
4DCT9 512 x 512 x 128 0.97 x 0.97 x 2.5 342 7.82 (3.99) 150/3 0.75 (1.09)
4DCT10 512 x 512 x 120 0.97 x 0.97 x 2.5 435 7.63 (6.54) 150/3 0.86 (1.45)
Table 1b. Reference Datasets - COPDgene. Each case is identified according to the given label. The image dimensions are given in voxel units, and the voxel dimensions are given in millimeters. The "# Features" column designates the total quantity of unique landmark features identified for each case. For the 4DCT images, the full point set is identified between the maximum inhalation and exhalation component phase images. Additionally, a subset of 75 features has been propagated onto each of the expiratory phase images (i.e., T00, T10, T20, T30, T40, and T50). The "Displacement" column shows the mean (and standard deviation) displacement of the complete primary feature set.The entries in the "# Repeats" column are formatted as (Nm / Nobs), where Nm is the number of repeat registration measurements performed by each of Nobs independent observers. The "Observers" column shows the combined mean (and standard deviation) repeat registration error for the set of Nobs data sets. Please see the references cited below for more information.
Label Image Dims Voxels (mm) # Features Displacement (mm) # Repeats Observers (mm)
COPD1 512 x 512 x 121 0.625 x 0.625 x 2.5 773 25.90 (11.57) 150/3 0.65 (0.73)
COPD2 512 x 512 x 102 0.645 x 0.645 x 2.5 612 21.77 (6.46) 150/3 1.06 (1.51)
COPD3 512 x 512 x 126 0.652 x 0.652 x 2.5 1172 12.29 (6.39) 150/3 0.58 (0.87)
COPD4 512 x 512 x 126 0.590 x 0.590 x 2.5 786 30.90 (13.49) 150/3 0.71 (0.96)
COPD5 512 x 512 x 131 0.647 x 0.647 x 2.5 1029 30.90 (14.05) 150/3 0.65 (0.87)
COPD6 512 x 512 x 119 0.633 x 0.633 x 2.5 633 28.32 (9.20) 150/3 1.06 (2.38)
COPD7 512 x 512 x 112 0.625 x 0.625 x 2.5 575 21.66 (7.66) 150/3 0.65 (0.78)
COPD8 512 x 512 x 115 0.586 x 0.586 x 2.5 791 25.57 (13.61) 150/3 0.96 (3.07)
COPD9 512 x 512 x 116 0.664 x 0.664 x 2.5 447 14.84 (10.01) 150/3 1.01 (2.54)
COPD10 512 x 512 x 135 0.742 x 0.742 x 2.5 480 22.48 (10.64) 150/3 0.87 (1.65)

Peer-Reviewed Publications

Links to the original peer-reviewed publications describing the data acquisition and analysis process for all of the reference datasets available through this website are provided below.

DIR-Lab case numbers 4DCT1-4DCT5:

Castillo R, Castillo E, Guerra R, Johnson VE, McPhail T, Garg AK, Guerrero T. 2009. A framework for evaluation of deformable image registration spatial accuracy using large landmark point sets. Phys Med Biol 54 1849-1870.

DIR-Lab case numbers 4DCT6-4DCT10:

Castillo E, Castillo R, Martinez J, Shenoy M, Guerrero T. 2009. Four-dimensional deformable image registration using trajectory modeling. Phys Med Biol 55 305-327.

DIR-Lab case numbers COPD1-COPD10:

Castillo R, Castillo E, Fuentes D, Ahmad M, Wood AM, Ludwig MS, Guerrero T. 2013. A reference dataset for deformable image registration spatial accuracy evaluation using the COPDgene study archive. Phys Med Biol 58 2861-2877.