SpineWeb

Collaborative Platform for Research on Spine Imaging and Image Analysis


edit

Datasets

Filed in: Main.Datasets · Modified on : Fri, 20 Jan 17

Dataset 15: Test set for CSI 2014 Vertebra Segmentation Challenge

This is the test data for the segmentation challenge of the CSI 2014 Workshop.

• The data sets contain 10 spine CTs acquired during daily clinical routine work in a trauma center at the Department of Radiological Sciences, University of California, Irvine, School of Medicine. Five data sets were from young adult (20-35 years old). The other five were vertebral compression fracture. • The scans cover the entire thoracic and lumbar spine. The data sets were acquired without intravenous contrast. The in-plane resolution is between 0.31 and 0.45mm and the slice thickness is 1mm or 2mm. • Image data are provided in MetaImage file format (mhd/raw). File name format is xxx.mhd.

This data is intended for research purposes, and is provided for non-commercial use only. By downloading the data, you accept the ODC Public Domain Dedication and License.

If you use the database for publication of any kind, please reference the following paper:

Jianhua Yao, Joseph E. Burns, Daniel Forsberg, Alexander Seitel, Abtin Rasoulian, Purang Abolmaesumi, Kerstin Hammernik, Martin Urschler, Bulat Ibragimov, Robert Korez, Tomaž Vrtovec, Isaac Castro-Mateos, Jose M. Pozo, Alejandro F. Frangi, Ronald M. Summers, Shuo Li, A Multi-center Milestone Study of Clinical Vertebral CT Segmentation, accepted for publication in Computerized Medical Imaging and Graphics, 2016.

To obtain, please click here

Dataset 14: IVDM3Seg – Intervertebral Disc (IVD) Localization and Segmentation from 3D Multi-Modality MR (M3) Images

An open online computational challenge in the field of spine imaging.

There are 24 3D multi-modality MRI data sets of at least 7 IVDs of the lower spine, collected from 12 subjects in two different stages in a study investigating the effect of prolonged bed rest (spaceflight simulation) on the lumbar intervertebral discs. Each subject at each stage was scanned with a 1.5-Tesla MRI scanner of Siemens using Dixon protocol. Thus, each 3D multi-modality MRI data set contains four aligned high-resolution 3D volumes: in-phase, opposed-phase, fat and water images. In total we will have 96 high resolution 3D MRI volume data. For each IVD, reference manual segmentation is provided in the form of binary mask. All images (four volumes per patient) and binary masks (one binary volume per patient) are stored in the Neuroimaging Informatics Technology Initiative (NIFTI) file format.

To obtain please contact Dr. Zheng directly. Email: guoyan.zheng@istb.unibe.ch

Dataset 13: xVertSeg - Segmentation and Classification of Fractured Vertebrae

An open online computational challenge in the field of spine medical image analysis

The xVertSeg challenge aims to bring together researchers who are interested in spinal imaging and image analysis, especially in vertebra segmentation and fracture classification problems, and validate the performance of their existing or newly developed algorithms on the same standard database of clinical images with vertebral fractures.

The xVertSeg.v1 database contains 25 CT lumbar spine images including non-fractured vertebrae and vertebrae with fractures of different morphological grades and cases. For 15 images, corresponding vertebra segmentation masks and fracture classification scores are provided, therefore these images should be used for research and/or training purposes. For 10 images, corresponding vertebra segmentation masks and fracture classification scores are known to challenge organizers only, therefore these images should be used for testing purposes, and will be used by challenge organizers for performance evaluation.

To obtain, click here

Dataset 12: Testing set A of the MICCAI Challenge on Vertebral Fracture Analysis

This dataset is the Testing Set A of the MICCAI 2015 Challenge “Automatic vertebral fracture analysis and identification from VFA by DXA”. It includes pairs of VFA images in two perpendicular views (lateral and anterior-posterior) for 30 subjects. They include approximately 10 subjects with compression fracture, 10 subjects with other deformations, and 10 subjects with normal shaped vertebrae.

A detailed description of the images is found here.

The Challenge is described here.

The terms and conditions of usage of this dataset are here

To obtain, please click here

Dataset 11: High anisotropy MRIs of the lower back

Lower back pain is affecting increasing number of people with increase in office work. Authoritative diagnosis is frequently established using highly anisotropic MRI scans. This poses unique challenges. The research to tackle this was conducted between 2009 and 2014. Program code is made publicly available under an open source license, including 17 anonymized datasets with corresponding manual segmentations.

Datasets came from variety of hospitals, with a high variety (T1, T2 and TIRM sequences, and a range of TE, TR and other parameters). More details about data can be found in the following publication:

Dženan Zukić, Aleš Vlasák, Jan Egger, Daniel Hořínek, Christopher Nimsky, Andreas Kolb - Robust Detection and Segmentation for Diagnosis of Vertebral Diseases using Routine MR Images. In Computer Graphics Forum, 33(6), 2014, pages 190-204.

If you use these datasets in your work, it would be courteous to cite the above paper.

Further information:

To obtain, please click here

Dataset 10: Multi-Modality Vertebra Recognition in Arbitrary Views using 3D Deformab le Hierarchical Model

Spine structure analysis requires the understanding of vertebra locations, poses and naming in both MR and CT images. The proposed dataset aims to provide a new data reference for the increasing studies of spine recognition problems. The dataset contains MR+CT images from 20 subjects. For each vertebra, the 3D vertebra center location (at spinal cord) and the vertebra 3D orientation are manually annotated as ground truth.

Any publication resulting from any use of this dataset may cite the following paper: Yunliang Cai, Said Osman, Manas Sharma, Mark Landis, and Shuo Li, “Multi-Modality Vertebra Recognition in Arbitrary Views using 3D Deformable Hierarchical Model”, IEEE Transactions on Medical Imaging, 2015.

Contact: Dr. Yunliang Cai (ycai82_at_uwo.ca) or Dr. Shuo Li (slishuo_at_gmail.com)

To obtain, please click this

Dataset 9: Automatic vertebral fracture analysis and identification from VFA by DXA

Osteoporotic vertebral fractures are difficult to diagnose and are often only discovered when the spine is imaged. A clinical assessment of vertebral fracture status can be performed using dual energy x-ray absorptiometry (DXA). This approach is known as vertebral fracture assessment (VFA). The datasets provided for the challenge consist of lateral and posterior-anterior (PA) scan images of the thoracolumbar spine acquired at a resolution of between 1 and 0.35 mm with Hologic Discovery A DXA scanner using the Instant Vertebral Assessment (IVA) scan option. Examples of these images (showing no vertebral fractures) are shown here (PA view (left), lateral view (right)).

In contrast with traumatic fractures, osteoporotic fractures typically affect the vertebral body. Depending on the fracture morphology, they can be classified into three types: concave, wedge and crush.

The major challenge encountered during the vertebral fracture assessment process is to determine whether deviations in the size and shape of vertebrae are due to true fractures or are simply non-fracture deformities, commonly observed in the general population.

The datasets include cases with osteoporotic vertebral fractures, cases with non-fracture deformities, and normal cases. This classification is included per vertebra and per patient.

More info here

To obtain please click here

Dataset 8: Osteoporotic vertebral fractures

The database consists of four CT lumbar spine images with osteoporotic vertebral fractures at one or more level (i.e. levels from L1 to L5). For each vertebra, reference manual segmentation is provided in the form of a binary mask. All images and binary masks are stored in the metaimage (MHD-RAW) format used by ITK.

The database is a preview of a larger database (of approx. 30 images) that is planned to be released by autumn 2015 as material for a tentative CSI 2016 challenge.

Any publication resulting from any use of this database must cite the following paper:

"Robert Korez, Bulat Ibragimov, Boštjan Likar, Franjo Pernuš, and Tomaž Vrtovec, “A framework for automated spine and vertebrae interpolation-based detection and model-based segmentation”, IEEE Transactions on Medical Imaging, in press, 2015 "

To obtain please click here

Contact: Robert Korez (Robert.korez@fe.uni-lj.si) or Tomaž Vrtovec (tomaz.vrtovec@fe.uni-lj.si) Laboratory of Imaging Technologies University of Ljubljana, Faculty of Electrical Engineering, Slovenia Official website: http://lit.fe.uni-lj.si

Dataset 7: Intervertebral Disc Localization and Segmentation: 3D T2-weighted Turbo Spin Echo MR image Database

The database consists of 3D T2-weighted turbo spin echo MR images of 15 anonymized patients, each containing at least 7 intervertebral discs (IVDs) of the lower spine (T11-L5, Fig. 1). Each patient was scanned with 1.5 Tesla MRI scanner of Siemens (Siemens Magnetom Sonata, Siemens Healthcare, Erlangen, Germany). For each IVD, reference manual segmentation is provided in the form of a binary mask. All images and binary masks (one binary volume per patient) are stored in the Neuroimaging Informatics Technology Initiative (NIFTI) file format.

Contact:

Guoyan Zheng, PD PhD
Institute for Surgical Technology and Biomechanics,
University of Bern,
Stauffacherstrasse 78,
CH-3014, Bern
Switzerland
Tel: +41-31-631-5956
Email: guoyan.zheng@istb.unibe.ch

To obtain please contact Dr. Zheng directly.

Dataset 6: Intervertebral Disc Localization and Segmentation Multi-modality MRI Spine Image Database

The database consists of multi-modality MRI spine images of 8 anonymized patients, each containing at least 7 intervertebral discs (IVDs) of the lower spine (T11 – L5). Each patient was scanned with 1.5 Tesla MRI scanner of Siemens. Dixon protocol was used to reconstruct four aligned high-resolution 3D volumes during one data acquisition (Fig. 1): in- phase, opposed-phase, fat and water images. For each IVD, reference manual segmentation is provided in the form of a binary mask. All images (four volumes per patient) and binary masks (one binary volume per patient) are stored in the Neuroimaging Informatics Technology Initiative (NIFTI) file format.

Fig. 1. The four aligned channels of a patient data (for visualization purpose, we only show the middle sagittal (mid-sagittal) slice of each channel.

Contact:

Guoyan Zheng, PD PhD Institute for Surgical Technology and Biomechanics, University of Bern, Stauffacherstrasse 78, CH-3014, Bern Switzerland

Tel: +41-31-631-5956 Email: guoyan.zheng@istb.unibe.ch

Dataset 5: Lumbar vertebra segmentation CT image database

The database consists of 10 CT lumbar spine images of 10 normal subjects, each containing 5 lumbar vertebrae (i.e. levels from L1 to L5). For each vertebra, reference manual segmentation is provided in the form of a binary mask. All images and binary masks are stored in the nearly raw raster data (NRRD) format.

Any publication resulting from any use of this database must cite the following paper:

Bulat Ibragimov, Boštjan Likar, Franjo Pernuš, and Tomaž Vrtovec, “Shape representation for efficient landmark-based segmentation in 3D”, IEEE Transactions on Medical Imaging, 33(4):861-874, 2014 [doi:10.1109/TMI.2013.2296976]

Contact:

Bulat Ibragimov (bulat.ibragimov@fe.uni-lj.si) or Tomaž Vrtovec (tomaz.vrtovec@fe.uni-lj.si)
Laboratory of Imaging Technologies
University of Ljubljana, Faculty of Electrical Engineering, Slovenia
Official website: http://lit.fe.uni-lj.si

Dataset 4: CVIP Spinal CT Database

The CVIP spinal CT database is provided for only research purposes.

All publications and studies that use any of the provided database must reference [1] or [2]:

[1]Melih S. Aslan, Ahmed Shalaby, and Aly A. Farag. "Clinically desired segmentation method for vertebral bodies." In Biomedical Imaging (ISBI), 2013 IEEE 10th International Symposium on, pp. 840-843. IEEE, 2013.

[2] Melih S. Aslan, Asem Ali, Ham Rara, Ben Arnold, Aly A. Farag, Rachid Fahmi, and Ping Xiang. "A novel 3d segmentation of vertebral bones from volumetric ct images using graph cuts." In Advances in Visual Computing, pp. 519-528. Springer Berlin Heidelberg, 2009.

Those who will need more would need contact with Dr. Aly A. Farag (aly.farag AT louisville DOT edu ) to fill a form for CVIP Lab.

What is given with this data:

  • CT scans of 5 patients (both dicom and bmp formats)
  • Ground truths: All scans have ground truths with original size (i.e., 512x512). In some data, if a VB is not fully scanned, we do not manually segment those half scans.
  • To obtain please click here

1 Anonbv 154_fbhs Nim=80;
2 Anonja_fbacks_25_64 Nim=64;
3 Anonsm 2355 Nim=96;
4 Ivy3_80 Nim=80;
5 MESAYUXXREM Nim=29;

Dataset 3: Vertebrae Localization and Identification

The database consists of spine-focused (i.e. tightly cropped) CT scans of 125 patients with varying types of pathologies. For most patients, multiple scans from longitudinal examinations are available, resulting in overall 242 scans in the database. For each scan, manual annotations of vertebrae centroids are provided. The data has been acquired at the Department of Radiology at the University of Washington.

This is the training data for the localization and identification challenge of the CSI 2014 Workshop.

Official website: http://research.microsoft.com/spine/
Contact: Ben Glocker (b.glocker@imperial.ac.uk)

Acknowledgments:
The data has been provided by the Department of Radiology at University of Washington. The public release has been supported by the Microsoft Research Connections Medical Imaging Initiative.

Dataset 2: Spine and Vertebrae Segmentation

Segmentation is an essential step for many computational spine imaging tasks. Clinical datasets raise many difficulties for automatic methods and ground truth is often scarce. To facilitate and foster the research on this topic, we provide a database of 10 spine CT scans along with manual segmentation of the thoracic and lumbar vertebrae. The data set can be used for development, training and evaluation of spine segmentation algorithm.

This is the training data for the segmentation challenge of the CSI 2014 Workshop.

  • To obtain please click here

Data Description

  • The data sets are spine CTs acquired during daily clinical routine work in a trauma center from 10 young adult (16-35 years old). The scans cover the entire thoracic and lumbar spine. The data sets were acquired without intravenous contrast. The in-plane resolution is between 0.31 and 0.45mm and the slice thickness is 1mm.
  • The data have been acquired at the Department of Radiological Sciences, University of California, Irvine, School of Medicine
  • Images were acquired with Philips or Siemens multidetector CT scanners.
  • Image data are provided in MetaImage file format (mhd/raw). File name format is xxx.mhd.

Annotations

  • In each scan, all complete thoracic and lumbar vertebrae have been semi-automatically segmented and verified.
  • The manual segmentation mask is stored in MetaImage file format (mhd/raw). Each vertebra is assigned a different label. First vertebra is labeled as 100, second as 200, and so on. File name is xxx_label.mhd.

Terms and Conditions

  • If you use the database for publication of any kind, please reference the following paper:
Jianhua Yao, Joseph E. Burns, Hector Muñoz and Ronald M. Summers, Detection of Vertebral Body Fractures Based on Cortical Shell Unwrapping, International Conference on Medical Image Computing and Computer Assisted Intervention, Volume 7512, 2012, pp 509-516
  • This data is intended for research purposes, and is provided for non-commercial use only. By downloading the data, you accept the ODC Public Domain Dedication and License (PDDL) (http://opendatacommons.org/licenses/pddl/1.0/).

People Involved

  • Jianhua Yao1 (primary contact, jyao@cc.nih.gov)
  • Joseph Burns2
  • Sasha Getty2
  • James Stieger1
  • Ronald Summers1

1. Radiology and Imaging Sciences Department, Clinical Center, National Institutes of Health, USA

2. Department of Radiological Sciences, University of California, Irvine, School of Medicine

Dataset 1: Cross Modality Spinal Images for Spine Workshop

Spine: 30 pairs of spinal CT and MR from lumber. Each pair of CT and MR is from same subject. Ground truth is not yet available.

  • To obtain please click here


Powered by PmWiki